Download validation.jsonl from HellGateSys/clef-orchestrator-docreview-sft: direct link, hf CLI and curl.
- Browser
- Download file 254 kB
-
https://huggingface.co/datasets/HellGateSys/clef-orchestrator-docreview-sft/resolve/main/validation.jsonl
- Command line
-
hf download hf://datasets/HellGateSys/clef-orchestrator-docreview-sft/validation.jsonl
-
curl -L -o validation.jsonl https://huggingface.co/datasets/HellGateSys/clef-orchestrator-docreview-sft/resolve/main/validation.jsonl
254 kB
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nUpon the expiration of the Term or termination of this Supply Agreement, in its entirety or with respect to one or more Products, this Supply Agreement shall, except as otherwise provided in this Section 10.3 or Section 10.5, be of no further force or effect; provided, however, that (a) in the event this Supply Agreement is terminated by SutroVax pursuant to Section 10.2.3 and there are outstanding Work Orders or other purchase orders accepted by Sutro that would not be fulfilled as a result..."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"clause flag check\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Upon the expiration of the Term or termination of this Supply Agreement, in its entirety or with respect to one or more Products, this Supply Agreement shall, except as otherwise provided in this Section 10.3 or Section 10.5, be of no further force or effect; provided, however, that (a) in the event this Supply Agreement is terminated by SutroVax pursuant to Section 10.2.3 and there are outstanding Work Orders or other purchase orders accepted by Sutro that would not be fulfilled as a result...\", \"questions\": {\"flag_post_termination_services\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain post-termination services clause language?\", \"criteria\": {\"true\": \"A post-termination services clause is present in the text.\", \"false\": \"No post-termination services clause language appears.\"}}}}}"}], "source": "cuad_packA"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nContract review against the standard 41-clause catalog.\n\nExcerpt:\nThis Limited Power of Attorney shall be revoked and terminated automatically upon the cancellation or termination of the Services Agreement or as to any Fund upon the cancellation or termination of the Adviser's Investment Advisory Contract for such Fund. ... FEDERATED ADVISORY SERVICES COMPANY"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Contract review against the standard 41-clause catalog.\\n\\nExcerpt:\\nThis Limited Power of Attorney shall be revoked and terminated automatically upon the cancellation or termination of the Services Agreement or as to any Fund upon the cancellation or termination of the Adviser's Investment Advisory Contract for such Fund. ... FEDERATED ADVISORY SERVICES COMPANY\", \"questions\": {\"flag_governing_law\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain governing law clause language?\", \"criteria\": {\"true\": \"A governing law clause is present in the text.\", \"false\": \"No governing law clause language appears.\"}}}}}"}], "source": "cuad_packB"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nContract review against the standard 41-clause catalog.\n\nExcerpt:\nThis Agreement will be in effect for three (3) years from the Effective Date (\"Initial Term\") unless terminated earlier in accordance with this Agreement. ... Zynga shall permit such records to be examined by authorized representatives of WPT, including such independent auditors as WPT may designate, during usual business hours, with advance notice, to verify to the extent necessary the Royalties paid hereunder, and WPT and its representatives shall use reasonable efforts to minimize disruptions to..."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Contract review against the standard 41-clause catalog.\\n\\nExcerpt:\\nThis Agreement will be in effect for three (3) years from the Effective Date (\\\"Initial Term\\\") unless terminated earlier in accordance with this Agreement. ... Zynga shall permit such records to be examined by authorized representatives of WPT, including such independent auditors as WPT may designate, during usual business hours, with advance notice, to verify to the extent necessary the Royalties paid hereunder, and WPT and its representatives shall use reasonable efforts to minimize disruptions to...\", \"questions\": {\"flag_audit_rights\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain audit rights clause language?\", \"criteria\": {\"true\": \"A audit rights clause is present in the text.\", \"false\": \"No audit rights clause language appears.\"}}}}}"}], "source": "cuad_packB"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nContract review against the standard 41-clause catalog.\n\nExcerpt:\nIn no event will Licensor be liable for any indirect, incidental, reliance, special or consequential damages arising out of the performance or non-performance of this Agreement, whether or not Licensor had been advised of the possibility of such damages. ... In consideration of the remuneration to be paid to Licensor pursuant hereto, Licensor grants to Company the right and license during the Contract Period to use the Duval Identification solely in connection with the advertisement and promotion of Company's Products within the Contract Territory as set forth herein."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Contract review against the standard 41-clause catalog.\\n\\nExcerpt:\\nIn no event will Licensor be liable for any indirect, incidental, reliance, special or consequential damages arising out of the performance or non-performance of this Agreement, whether or not Licensor had been advised of the possibility of such damages. ... In consideration of the remuneration to be paid to Licensor pursuant hereto, Licensor grants to Company the right and license during the Contract Period to use the Duval Identification solely in connection with the advertisement and promotion of Company's Products within the Contract Territory as set forth herein.\", \"questions\": {\"flag_cap_on_liability\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain cap on liability clause language?\", \"criteria\": {\"true\": \"A cap on liability clause is present in the text.\", \"false\": \"No cap on liability clause language appears.\"}}}}}"}], "source": "cuad_packB"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\ninstead of telling young women to dress appropriately, teach young men to NOT rape, that NO means NO and to love accordingly. #MeToo"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"hate speech screen\", \"request\": {\"model\": \"clef-omni\", \"state\": \"instead of telling young women to dress appropriately, teach young men to NOT rape, that NO means NO and to love accordingly. #MeToo\", \"questions\": {\"is_hate_speech\": {\"type\": \"noul\", \"instructions\": \"Does this message contain hate speech against a group?\", \"criteria\": {\"true\": \"Targets a protected group with hate.\", \"false\": \"No hate toward a protected group.\"}}}}}"}], "source": "hate"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation (JSON state):\n{\"subject\": \"Urgent Update Required for Enhanced Performance\", \"message\": \"Customer support, we need an urgent update to improve the scalability and performance of the affected products, such as 3D-Drucker, PC-Gehäuse, and Lenovo ThinkPad X1 Carbon. The update is crucial to ensure these products continue to function efficiently and effectively. We would greatly appreciate it if you could address this matter as soon as possible. Thank you for your attention to this issue.\"}"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"support ticket triage\", \"request\": {\"model\": \"clef-omni\", \"state\": {\"subject\": \"Urgent Update Required for Enhanced Performance\", \"message\": \"Customer support, we need an urgent update to improve the scalability and performance of the affected products, such as 3D-Drucker, PC-Gehäuse, and Lenovo ThinkPad X1 Carbon. The update is crucial to ensure these products continue to function efficiently and effectively. We would greatly appreciate it if you could address this matter as soon as possible. Thank you for your attention to this issue.\"}, \"questions\": {\"route_to\": {\"type\": \"choice\", \"instructions\": \"Which support queue should handle this ticket?\", \"criteria\": {\"IT & Technology/Hardware Support\": \"IT & Technology/Hardware Support\", \"Customer Service\": \"Customer Service\", \"Billing and Payments\": \"Billing and Payments\", \"People & Society/Culture & Society\": \"People & Society/Culture & Society\", \"Technical Support\": \"Technical Support\"}}, \"ticket_type\": {\"type\": \"choice\", \"instructions\": \"What kind of ticket is this?\", \"criteria\": {\"Request\": \"Request\", \"Incident\": \"Incident\", \"Change\": \"Change\", \"Problem\": \"Problem\"}}, \"priority\": {\"type\": \"score\", \"instructions\": \"How urgent is this ticket?\", \"criteria\": [\"Low, can wait days\", \"Medium, handle this week\", \"High, handle today\", \"Critical, service blocked\"]}, \"needs_escalation\": {\"type\": \"noul\", \"instructions\": \"Should this be escalated beyond first-line support?\"}}}}"}], "source": "tickets"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nMay 29, 2018 ... Sutro shall allocate its available Components and manufacturing capacity to provide SutroVax with quantities of such Product at least equal to the greater of (a) [***] of the amount of Product (or products equivalent to Product) that Sutro allocates for itself and its Affiliates (but in no event less than [***] liters of Extract per month and the..."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"May 29, 2018 ... Sutro shall allocate its available Components and manufacturing capacity to provide SutroVax with quantities of such Product at least equal to the greater of (a) [***] of the amount of Product (or products equivalent to Product) that Sutro allocates for itself and its Affiliates (but in no event less than [***] liters of Extract per month and the...\", \"questions\": {\"flag_governing_law\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain governing law clause language?\", \"criteria\": {\"true\": \"A governing law clause is present in the text.\", \"false\": \"No governing law clause language appears.\"}}, \"flag_change_of_control\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain change of control clause language?\", \"criteria\": {\"true\": \"A change of control clause is present in the text.\", \"false\": \"No change of control clause language appears.\"}}, \"flag_parties\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain parties clause language?\", \"criteria\": {\"true\": \"A parties clause is present in the text.\", \"false\": \"No parties clause language appears.\"}}, \"flag_audit_rights\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain audit rights clause language?\", \"criteria\": {\"true\": \"A audit rights clause is present in the text.\", \"false\": \"No audit rights clause language appears.\"}}, \"flag_competitive_restriction_exception\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain competitive restriction exception clause language?\", \"criteria\": {\"true\": \"A competitive restriction exception clause is present in the text.\", \"false\": \"No competitive restriction exception clause language appears.\"}}, \"flag_termination_for_convenience\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain termination for convenience clause language?\", \"criteria\": {\"true\": \"A termination for convenience clause is present in the text.\", \"false\": \"No termination for convenience clause language appears.\"}}, \"flag_insurance\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain insurance clause language?\", \"criteria\": {\"true\": \"A insurance clause is present in the text.\", \"false\": \"No insurance clause language appears.\"}}, \"flag_post_termination_services\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain post-termination services clause language?\", \"criteria\": {\"true\": \"A post-termination services clause is present in the text.\", \"false\": \"No post-termination services clause language appears.\"}}}}}"}], "source": "cuad_packD"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nI am still waiting for a transfer I did"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"I am still waiting for a transfer I did\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"order_physical_card\": \"order physical card\", \"automatic_top_up\": \"automatic top up\", \"balance_not_updated_after_bank_transfer\": \"balance not updated after bank transfer\", \"card_about_to_expire\": \"card about to expire\", \"compromised_card\": \"compromised card\", \"pending_top_up\": \"pending top up\", \"lost_or_stolen_card\": \"lost or stolen card\", \"virtual_card_not_working\": \"virtual card not working\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nDecember 14, 2015 ... TERM: November 01, 2014 to October 31, 2017"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"December 14, 2015 ... TERM: November 01, 2014 to October 31, 2017\", \"questions\": {\"flag_governing_law\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain governing law clause language?\", \"criteria\": {\"true\": \"A governing law clause is present in the text.\", \"false\": \"No governing law clause language appears.\"}}, \"flag_parties\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain parties clause language?\", \"criteria\": {\"true\": \"A parties clause is present in the text.\", \"false\": \"No parties clause language appears.\"}}, \"flag_insurance\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain insurance clause language?\", \"criteria\": {\"true\": \"A insurance clause is present in the text.\", \"false\": \"No insurance clause language appears.\"}}, \"flag_non_transferable_license\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain non-transferable license clause language?\", \"criteria\": {\"true\": \"A non-transferable license clause is present in the text.\", \"false\": \"No non-transferable license clause language appears.\"}}, \"flag_agreement_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain agreement date clause language?\", \"criteria\": {\"true\": \"A agreement date clause is present in the text.\", \"false\": \"No agreement date clause language appears.\"}}}}}"}], "source": "cuad_packD"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation (JSON state):\n{\"subject\": \"Support for Unplanned Downtime and Performance Issues\", \"message\": \"We have encountered unexpected outages and slow performance on the site, which may be due to server overload or configuration issues. Initial troubleshooting involved restarting affected services and reviewing logs, and resources have been scaled up, but the issue still persists.\"}"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"support ticket triage\", \"request\": {\"model\": \"clef-omni\", \"state\": {\"subject\": \"Support for Unplanned Downtime and Performance Issues\", \"message\": \"We have encountered unexpected outages and slow performance on the site, which may be due to server overload or configuration issues. Initial troubleshooting involved restarting affected services and reviewing logs, and resources have been scaled up, but the issue still persists.\"}, \"questions\": {\"route_to\": {\"type\": \"choice\", \"instructions\": \"Which support queue should handle this ticket?\", \"criteria\": {\"Customer Service\": \"Customer Service\", \"IT Support\": \"IT Support\", \"Health/Medical Services\": \"Health/Medical Services\", \"IT & Technology/Hardware Support\": \"IT & Technology/Hardware Support\", \"Sales and Pre-Sales\": \"Sales and Pre-Sales\"}}, \"ticket_type\": {\"type\": \"choice\", \"instructions\": \"What kind of ticket is this?\", \"criteria\": {\"Problem\": \"Problem\", \"Change\": \"Change\", \"Request\": \"Request\", \"Incident\": \"Incident\"}}, \"priority\": {\"type\": \"score\", \"instructions\": \"How urgent is this ticket?\", \"criteria\": [\"Low, can wait days\", \"Medium, handle this week\", \"High, handle today\", \"Critical, service blocked\"]}, \"needs_escalation\": {\"type\": \"noul\", \"instructions\": \"Should this be escalated beyond first-line support?\"}}}}"}], "source": "tickets"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\n\"No\" is the complete sentence I will use for this exercise."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"emotion tagging\", \"request\": {\"model\": \"clef-omni\", \"state\": \"\\\"No\\\" is the complete sentence I will use for this exercise.\", \"questions\": {\"emotion_disapproval\": {\"type\": \"noul\", \"instructions\": \"Does the author express disapproval?\"}}}}"}], "source": "go_emotions"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nLifeUSA/ Envision Health, Inc. ... This Agreement is deemed to have been entered into in the State of Colorado, and its interpretation, construction, and the remedies for its enforcement or breach are to be applied pursuant to and in accordance with the laws of the State of Colorado."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"LifeUSA/ Envision Health, Inc. ... This Agreement is deemed to have been entered into in the State of Colorado, and its interpretation, construction, and the remedies for its enforcement or breach are to be applied pursuant to and in accordance with the laws of the State of Colorado.\", \"questions\": {\"flag_minimum_commitment\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain minimum commitment clause language?\", \"criteria\": {\"true\": \"A minimum commitment clause is present in the text.\", \"false\": \"No minimum commitment clause language appears.\"}}, \"flag_exclusivity\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain exclusivity clause language?\", \"criteria\": {\"true\": \"A exclusivity clause is present in the text.\", \"false\": \"No exclusivity clause language appears.\"}}, \"flag_parties\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain parties clause language?\", \"criteria\": {\"true\": \"A parties clause is present in the text.\", \"false\": \"No parties clause language appears.\"}}, \"flag_effective_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain effective date clause language?\", \"criteria\": {\"true\": \"A effective date clause is present in the text.\", \"false\": \"No effective date clause language appears.\"}}, \"flag_rofr_rofo_rofn\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain rofr/rofo/rofn clause language?\", \"criteria\": {\"true\": \"A rofr/rofo/rofn clause is present in the text.\", \"false\": \"No rofr/rofo/rofn clause language appears.\"}}}}}"}], "source": "cuad_packD"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation (JSON state):\n{\"subject\": \"Request for Updated Hardware and Software Compatibility\", \"message\": \"Dear Customer Support Team, I am contacting you to inquire about the integration of new hardware and software compatibility to enhance our project management experience. As a valued customer, I believe it is crucial to adopt the latest technologies in order to achieve efficient and effective project management. The current system has some limitations, and I am requesting an update to include the latest hardware and software compatibility, which would facilitate better project management. Could you please share with me the current status of this integration and provide an estimated timeline for its implementation? Furthermore, are there any plans to introduce new features or updates that...\"}"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"support ticket triage\", \"request\": {\"model\": \"clef-omni\", \"state\": {\"subject\": \"Request for Updated Hardware and Software Compatibility\", \"message\": \"Dear Customer Support Team, I am contacting you to inquire about the integration of new hardware and software compatibility to enhance our project management experience. As a valued customer, I believe it is crucial to adopt the latest technologies in order to achieve efficient and effective project management. The current system has some limitations, and I am requesting an update to include the latest hardware and software compatibility, which would facilitate better project management. Could you please share with me the current status of this integration and provide an estimated timeline for its implementation? Furthermore, are there any plans to introduce new features or updates that...\"}, \"questions\": {\"route_to\": {\"type\": \"choice\", \"instructions\": \"Which support queue should handle this ticket?\", \"criteria\": {\"Science/Research\": \"Science/Research\", \"IT Support\": \"IT Support\", \"Human Resources\": \"Human Resources\", \"Product Support\": \"Product Support\", \"Sports\": \"Sports\"}}, \"ticket_type\": {\"type\": \"choice\", \"instructions\": \"What kind of ticket is this?\", \"criteria\": {\"Incident\": \"Incident\", \"Request\": \"Request\", \"Change\": \"Change\", \"Problem\": \"Problem\"}}, \"priority\": {\"type\": \"score\", \"instructions\": \"How urgent is this ticket?\", \"criteria\": [\"Low, can wait days\", \"Medium, handle this week\", \"High, handle today\", \"Critical, service blocked\"]}, \"needs_escalation\": {\"type\": \"noul\", \"instructions\": \"Should this be escalated beyond first-line support?\"}}}}"}], "source": "tickets"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nThe term of this Agreement shall be ten (10) years (the \"Term\") which shall commence on the date upon which the Company delivers to Distributor the last Sample, as defined hereinafter."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"clause flag check\", \"request\": {\"model\": \"clef-omni\", \"state\": \"The term of this Agreement shall be ten (10) years (the \\\"Term\\\") which shall commence on the date upon which the Company delivers to Distributor the last Sample, as defined hereinafter.\", \"questions\": {\"flag_effective_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain effective date clause language?\", \"criteria\": {\"true\": \"A effective date clause is present in the text.\", \"false\": \"No effective date clause language appears.\"}}}}}"}], "source": "cuad_packA"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nThis Limited Power of Attorney shall be revoked and terminated automatically upon the cancellation or termination of the Services Agreement or as to any Fund upon the cancellation or termination of the Adviser's Investment Advisory Contract for such Fund. ... FEDERATED ADVISORY SERVICES COMPANY"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"This Limited Power of Attorney shall be revoked and terminated automatically upon the cancellation or termination of the Services Agreement or as to any Fund upon the cancellation or termination of the Adviser's Investment Advisory Contract for such Fund. ... FEDERATED ADVISORY SERVICES COMPANY\", \"questions\": {\"flag_governing_law\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain governing law clause language?\", \"criteria\": {\"true\": \"A governing law clause is present in the text.\", \"false\": \"No governing law clause language appears.\"}}, \"flag_competitive_restriction_exception\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain competitive restriction exception clause language?\", \"criteria\": {\"true\": \"A competitive restriction exception clause is present in the text.\", \"false\": \"No competitive restriction exception clause language appears.\"}}, \"flag_anti_assignment\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain anti-assignment clause language?\", \"criteria\": {\"true\": \"A anti-assignment clause is present in the text.\", \"false\": \"No anti-assignment clause language appears.\"}}, \"flag_expiration_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain expiration date clause language?\", \"criteria\": {\"true\": \"A expiration date clause is present in the text.\", \"false\": \"No expiration date clause language appears.\"}}, \"flag_effective_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain effective date clause language?\", \"criteria\": {\"true\": \"A effective date clause is present in the text.\", \"false\": \"No effective date clause language appears.\"}}, \"flag_document_name\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain document name clause language?\", \"criteria\": {\"true\": \"A document name clause is present in the text.\", \"false\": \"No document name clause language appears.\"}}, \"flag_unlimited_all_you_can_eat_license\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain unlimited/all-you-can-eat-license clause language?\", \"criteria\": {\"true\": \"A unlimited/all-you-can-eat-license clause is present in the text.\", \"false\": \"No unlimited/all-you-can-eat-license clause language appears.\"}}}}}"}], "source": "cuad_packC"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nIn no event will Licensor be liable for any indirect, incidental, reliance, special or consequential damages arising out of the performance or non-performance of this Agreement, whether or not Licensor had been advised of the possibility of such damages. ... In consideration of the remuneration to be paid to Licensor pursuant hereto, Licensor grants to Company the right and license during the Contract Period to use the Duval Identification solely in connection with the advertisement and promotion of Company's Products within the Contract Territory as set forth herein."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"In no event will Licensor be liable for any indirect, incidental, reliance, special or consequential damages arising out of the performance or non-performance of this Agreement, whether or not Licensor had been advised of the possibility of such damages. ... In consideration of the remuneration to be paid to Licensor pursuant hereto, Licensor grants to Company the right and license during the Contract Period to use the Duval Identification solely in connection with the advertisement and promotion of Company's Products within the Contract Territory as set forth herein.\", \"questions\": {\"flag_warranty_duration\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain warranty duration clause language?\", \"criteria\": {\"true\": \"A warranty duration clause is present in the text.\", \"false\": \"No warranty duration clause language appears.\"}}, \"flag_exclusivity\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain exclusivity clause language?\", \"criteria\": {\"true\": \"A exclusivity clause is present in the text.\", \"false\": \"No exclusivity clause language appears.\"}}, \"flag_anti_assignment\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain anti-assignment clause language?\", \"criteria\": {\"true\": \"A anti-assignment clause is present in the text.\", \"false\": \"No anti-assignment clause language appears.\"}}, \"flag_effective_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain effective date clause language?\", \"criteria\": {\"true\": \"A effective date clause is present in the text.\", \"false\": \"No effective date clause language appears.\"}}, \"flag_cap_on_liability\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain cap on liability clause language?\", \"criteria\": {\"true\": \"A cap on liability clause is present in the text.\", \"false\": \"No cap on liability clause language appears.\"}}}}}"}], "source": "cuad_packD"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nHelp me cancel a transfer."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Help me cancel a transfer.\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"disposable_card_limits\": \"disposable card limits\", \"cancel_transfer\": \"cancel transfer\", \"terminate_account\": \"terminate account\", \"card_payment_wrong_exchange_rate\": \"card payment wrong exchange rate\", \"failed_transfer\": \"failed transfer\", \"get_physical_card\": \"get physical card\", \"exchange_charge\": \"exchange charge\", \"atm_support\": \"atm support\", \"cash_withdrawal_not_recognised\": \"cash withdrawal not recognised\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nIs the Workman vs. LP game this Friday?"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"sentiment rating\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Is the Workman vs. LP game this Friday?\", \"questions\": {\"sentiment\": {\"type\": \"score\", \"instructions\": \"How does the author feel in this message?\", \"criteria\": [\"Negative tone\", \"Neutral, no strong tone\", \"Positive tone\"]}, \"is_complaint\": {\"type\": \"noul\", \"instructions\": \"Is the author complaining about a product or service?\"}}}}"}], "source": "sentiment"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nDuring the Term of this Agreement and for a period of twelve (12) months thereafter, the Distributor (on behalf of itself, each of its affiliates and each of their respective representatives) agrees that it will not directly or indirectly solicit or hire any executive, managerial or technical employee of the Company or any of its affiliates. ... If Distributor complies with all of the terms of this Agreement, the Agreement shall be renewable on an annual basis for one (1) year terms for up to another ten (10) years on the same terms and conditions as set forth herein."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"During the Term of this Agreement and for a period of twelve (12) months thereafter, the Distributor (on behalf of itself, each of its affiliates and each of their respective representatives) agrees that it will not directly or indirectly solicit or hire any executive, managerial or technical employee of the Company or any of its affiliates. ... If Distributor complies with all of the terms of this Agreement, the Agreement shall be renewable on an annual basis for one (1) year terms for up to another ten (10) years on the same terms and conditions as set forth herein.\", \"questions\": {\"flag_effective_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain effective date clause language?\", \"criteria\": {\"true\": \"A effective date clause is present in the text.\", \"false\": \"No effective date clause language appears.\"}}, \"flag_governing_law\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain governing law clause language?\", \"criteria\": {\"true\": \"A governing law clause is present in the text.\", \"false\": \"No governing law clause language appears.\"}}, \"flag_affiliate_license_licensor\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain affiliate license-licensor clause language?\", \"criteria\": {\"true\": \"A affiliate license-licensor clause is present in the text.\", \"false\": \"No affiliate license-licensor clause language appears.\"}}, \"flag_insurance\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain insurance clause language?\", \"criteria\": {\"true\": \"A insurance clause is present in the text.\", \"false\": \"No insurance clause language appears.\"}}, \"flag_renewal_term\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain renewal term clause language?\", \"criteria\": {\"true\": \"A renewal term clause is present in the text.\", \"false\": \"No renewal term clause language appears.\"}}, \"flag_agreement_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain agreement date clause language?\", \"criteria\": {\"true\": \"A agreement date clause is present in the text.\", \"false\": \"No agreement date clause language appears.\"}}, \"flag_notice_period_to_terminate_renewal\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain notice period to terminate renewal clause language?\", \"criteria\": {\"true\": \"A notice period to terminate renewal clause is present in the text.\", \"false\": \"No notice period to terminate renewal clause language appears.\"}}, \"flag_post_termination_services\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain post-termination services clause language?\", \"criteria\": {\"true\": \"A post-termination services clause is present in the text.\", \"false\": \"No post-termination services clause language appears.\"}}}}}"}], "source": "cuad_packC"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nOnly I could get kicked out of a club for just being"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"hate speech screen\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Only I could get kicked out of a club for just being\", \"questions\": {\"is_hate_speech\": {\"type\": \"noul\", \"instructions\": \"Does this message contain hate speech against a group?\", \"criteria\": {\"true\": \"Targets a protected group with hate.\", \"false\": \"No hate toward a protected group.\"}}}}}"}], "source": "hate"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\n\"Shawn's a smart guy, he may be young but he knows what he's talking about.\""}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"sentiment rating\", \"request\": {\"model\": \"clef-omni\", \"state\": \"\\\"Shawn's a smart guy, he may be young but he knows what he's talking about.\\\"\", \"questions\": {\"sentiment\": {\"type\": \"score\", \"instructions\": \"How does the author feel in this message?\", \"criteria\": [\"Negative tone\", \"Neutral, no strong tone\", \"Positive tone\"]}, \"is_complaint\": {\"type\": \"noul\", \"instructions\": \"Is the author complaining about a product or service?\"}}}}"}], "source": "sentiment"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nMy account shows a direct debit charge that I have never seen before"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"My account shows a direct debit charge that I have never seen before\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"transfer_fee_charged\": \"transfer fee charged\", \"card_payment_not_recognised\": \"card payment not recognised\", \"contactless_not_working\": \"contactless not working\", \"direct_debit_payment_not_recognised\": \"direct debit payment not recognised\", \"card_not_working\": \"card not working\", \"cash_withdrawal_charge\": \"cash withdrawal charge\", \"compromised_card\": \"compromised card\", \"topping_up_by_card\": \"topping up by card\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nI miss u bUT ALSO FUCK U, CUNT WHORE BITCH"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"hate speech screen\", \"request\": {\"model\": \"clef-omni\", \"state\": \"I miss u bUT ALSO FUCK U, CUNT WHORE BITCH\", \"questions\": {\"is_hate_speech\": {\"type\": \"noul\", \"instructions\": \"Does this message contain hate speech against a group?\", \"criteria\": {\"true\": \"Targets a protected group with hate.\", \"false\": \"No hate toward a protected group.\"}}}}}"}], "source": "hate"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nSERVICES AGREEMENT"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"clause flag check\", \"request\": {\"model\": \"clef-omni\", \"state\": \"SERVICES AGREEMENT\", \"questions\": {\"flag_document_name\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain document name clause language?\", \"criteria\": {\"true\": \"A document name clause is present in the text.\", \"false\": \"No document name clause language appears.\"}}}}}"}], "source": "cuad_packA"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nInsurance covers stupid."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"emotion tagging\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Insurance covers stupid.\", \"questions\": {\"emotion_annoyance\": {\"type\": \"noul\", \"instructions\": \"Does the author express annoyance?\"}}}}"}], "source": "go_emotions"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nWhy did my payment not go through why did it cancel?"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Why did my payment not go through why did it cancel?\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"top_up_reverted\": \"top up reverted\", \"top_up_by_bank_transfer_charge\": \"top up by bank transfer charge\", \"cancel_transfer\": \"cancel transfer\", \"reverted_card_payment?\": \"reverted card payment?\", \"card_payment_fee_charged\": \"card payment fee charged\", \"fiat_currency_support\": \"fiat currency support\", \"pin_blocked\": \"pin blocked\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nDecember 14, 2015"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"clause flag check\", \"request\": {\"model\": \"clef-omni\", \"state\": \"December 14, 2015\", \"questions\": {\"flag_agreement_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain agreement date clause language?\", \"criteria\": {\"true\": \"A agreement date clause is present in the text.\", \"false\": \"No agreement date clause language appears.\"}}}}}"}], "source": "cuad_packA"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nThis Limited Power of Attorney shall be revoked and terminated automatically upon the cancellation or termination of the Services Agreement or as to any Fund upon the cancellation or termination of the Adviser's Investment Advisory Contract for such Fund. ... FEDERATED ADVISORY SERVICES COMPANY"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"This Limited Power of Attorney shall be revoked and terminated automatically upon the cancellation or termination of the Services Agreement or as to any Fund upon the cancellation or termination of the Adviser's Investment Advisory Contract for such Fund. ... FEDERATED ADVISORY SERVICES COMPANY\", \"questions\": {\"flag_competitive_restriction_exception\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain competitive restriction exception clause language?\", \"criteria\": {\"true\": \"A competitive restriction exception clause is present in the text.\", \"false\": \"No competitive restriction exception clause language appears.\"}}, \"flag_anti_assignment\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain anti-assignment clause language?\", \"criteria\": {\"true\": \"A anti-assignment clause is present in the text.\", \"false\": \"No anti-assignment clause language appears.\"}}, \"flag_expiration_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain expiration date clause language?\", \"criteria\": {\"true\": \"A expiration date clause is present in the text.\", \"false\": \"No expiration date clause language appears.\"}}, \"flag_effective_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain effective date clause language?\", \"criteria\": {\"true\": \"A effective date clause is present in the text.\", \"false\": \"No effective date clause language appears.\"}}, \"flag_document_name\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain document name clause language?\", \"criteria\": {\"true\": \"A document name clause is present in the text.\", \"false\": \"No document name clause language appears.\"}}, \"flag_unlimited_all_you_can_eat_license\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain unlimited/all-you-can-eat-license clause language?\", \"criteria\": {\"true\": \"A unlimited/all-you-can-eat-license clause is present in the text.\", \"false\": \"No unlimited/all-you-can-eat-license clause language appears.\"}}, \"flag_governing_law\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain governing law clause language?\", \"criteria\": {\"true\": \"A governing law clause is present in the text.\", \"false\": \"No governing law clause language appears.\"}}}}}"}], "source": "cuad_packD"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nYoung Woman Gets Loud While She Stuffs Her Cunt"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"hate speech screen\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Young Woman Gets Loud While She Stuffs Her Cunt\", \"questions\": {\"is_hate_speech\": {\"type\": \"noul\", \"instructions\": \"Does this message contain hate speech against a group?\", \"criteria\": {\"true\": \"Targets a protected group with hate.\", \"false\": \"No hate toward a protected group.\"}}}}}"}], "source": "hate"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nHow can I exchange USD and GBP in the app?"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"How can I exchange USD and GBP in the app?\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"age_limit\": \"age limit\", \"request_refund\": \"request refund\", \"failed_transfer\": \"failed transfer\", \"pending_transfer\": \"pending transfer\", \"contactless_not_working\": \"contactless not working\", \"card_arrival\": \"card arrival\", \"get_disposable_virtual_card\": \"get disposable virtual card\", \"exchange_via_app\": \"exchange via app\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nUpon termination, ENVISION shall have eighteen (18) months to exhaust any inventories, packaging and advertising materials bearing the \"SierraSil\" trademark and SIERRA shall have first option to buy back any inventory at ENVISION's net purchase price."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"clause flag check\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Upon termination, ENVISION shall have eighteen (18) months to exhaust any inventories, packaging and advertising materials bearing the \\\"SierraSil\\\" trademark and SIERRA shall have first option to buy back any inventory at ENVISION's net purchase price.\", \"questions\": {\"flag_rofr_rofo_rofn\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain rofr/rofo/rofn clause language?\", \"criteria\": {\"true\": \"A rofr/rofo/rofn clause is present in the text.\", \"false\": \"No rofr/rofo/rofn clause language appears.\"}}}}}"}], "source": "cuad_packA"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nIn no event will Licensor be liable for any indirect, incidental, reliance, special or consequential damages arising out of the performance or non-performance of this Agreement, whether or not Licensor had been advised of the possibility of such damages. ... In consideration of the remuneration to be paid to Licensor pursuant hereto, Licensor grants to Company the right and license during the Contract Period to use the Duval Identification solely in connection with the advertisement and promotion of Company's Products within the Contract Territory as set forth herein."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"In no event will Licensor be liable for any indirect, incidental, reliance, special or consequential damages arising out of the performance or non-performance of this Agreement, whether or not Licensor had been advised of the possibility of such damages. ... In consideration of the remuneration to be paid to Licensor pursuant hereto, Licensor grants to Company the right and license during the Contract Period to use the Duval Identification solely in connection with the advertisement and promotion of Company's Products within the Contract Territory as set forth herein.\", \"questions\": {\"flag_cap_on_liability\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain cap on liability clause language?\", \"criteria\": {\"true\": \"A cap on liability clause is present in the text.\", \"false\": \"No cap on liability clause language appears.\"}}, \"flag_warranty_duration\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain warranty duration clause language?\", \"criteria\": {\"true\": \"A warranty duration clause is present in the text.\", \"false\": \"No warranty duration clause language appears.\"}}, \"flag_exclusivity\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain exclusivity clause language?\", \"criteria\": {\"true\": \"A exclusivity clause is present in the text.\", \"false\": \"No exclusivity clause language appears.\"}}, \"flag_anti_assignment\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain anti-assignment clause language?\", \"criteria\": {\"true\": \"A anti-assignment clause is present in the text.\", \"false\": \"No anti-assignment clause language appears.\"}}, \"flag_effective_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain effective date clause language?\", \"criteria\": {\"true\": \"A effective date clause is present in the text.\", \"false\": \"No effective date clause language appears.\"}}}}}"}], "source": "cuad_packC"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nDuring the Term of this Agreement and for a period of twelve (12) months thereafter, the Distributor (on behalf of itself, each of its affiliates and each of their respective representatives) agrees that it will not directly or indirectly solicit or hire any executive, managerial or technical employee of the Company or any of its affiliates. ... If Distributor complies with all of the terms of this Agreement, the Agreement shall be renewable on an annual basis for one (1) year terms for up to another ten (10) years on the same terms and conditions as set forth herein."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"During the Term of this Agreement and for a period of twelve (12) months thereafter, the Distributor (on behalf of itself, each of its affiliates and each of their respective representatives) agrees that it will not directly or indirectly solicit or hire any executive, managerial or technical employee of the Company or any of its affiliates. ... If Distributor complies with all of the terms of this Agreement, the Agreement shall be renewable on an annual basis for one (1) year terms for up to another ten (10) years on the same terms and conditions as set forth herein.\", \"questions\": {\"flag_governing_law\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain governing law clause language?\", \"criteria\": {\"true\": \"A governing law clause is present in the text.\", \"false\": \"No governing law clause language appears.\"}}, \"flag_affiliate_license_licensor\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain affiliate license-licensor clause language?\", \"criteria\": {\"true\": \"A affiliate license-licensor clause is present in the text.\", \"false\": \"No affiliate license-licensor clause language appears.\"}}, \"flag_insurance\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain insurance clause language?\", \"criteria\": {\"true\": \"A insurance clause is present in the text.\", \"false\": \"No insurance clause language appears.\"}}, \"flag_renewal_term\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain renewal term clause language?\", \"criteria\": {\"true\": \"A renewal term clause is present in the text.\", \"false\": \"No renewal term clause language appears.\"}}, \"flag_agreement_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain agreement date clause language?\", \"criteria\": {\"true\": \"A agreement date clause is present in the text.\", \"false\": \"No agreement date clause language appears.\"}}, \"flag_notice_period_to_terminate_renewal\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain notice period to terminate renewal clause language?\", \"criteria\": {\"true\": \"A notice period to terminate renewal clause is present in the text.\", \"false\": \"No notice period to terminate renewal clause language appears.\"}}, \"flag_post_termination_services\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain post-termination services clause language?\", \"criteria\": {\"true\": \"A post-termination services clause is present in the text.\", \"false\": \"No post-termination services clause language appears.\"}}, \"flag_effective_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain effective date clause language?\", \"criteria\": {\"true\": \"A effective date clause is present in the text.\", \"false\": \"No effective date clause language appears.\"}}}}}"}], "source": "cuad_packD"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nMy refund for a purchase hasn't cone through yet."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"My refund for a purchase hasn't cone through yet.\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"failed_transfer\": \"failed transfer\", \"get_disposable_virtual_card\": \"get disposable virtual card\", \"age_limit\": \"age limit\", \"Refund_not_showing_up\": \"Refund not showing up\", \"declined_cash_withdrawal\": \"declined cash withdrawal\", \"exchange_charge\": \"exchange charge\", \"exchange_rate\": \"exchange rate\", \"balance_not_updated_after_bank_transfer\": \"balance not updated after bank transfer\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nLifeUSA/ Envision Health, Inc. ... This Agreement is deemed to have been entered into in the State of Colorado, and its interpretation, construction, and the remedies for its enforcement or breach are to be applied pursuant to and in accordance with the laws of the State of Colorado."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"LifeUSA/ Envision Health, Inc. ... This Agreement is deemed to have been entered into in the State of Colorado, and its interpretation, construction, and the remedies for its enforcement or breach are to be applied pursuant to and in accordance with the laws of the State of Colorado.\", \"questions\": {\"flag_rofr_rofo_rofn\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain rofr/rofo/rofn clause language?\", \"criteria\": {\"true\": \"A rofr/rofo/rofn clause is present in the text.\", \"false\": \"No rofr/rofo/rofn clause language appears.\"}}, \"flag_minimum_commitment\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain minimum commitment clause language?\", \"criteria\": {\"true\": \"A minimum commitment clause is present in the text.\", \"false\": \"No minimum commitment clause language appears.\"}}, \"flag_exclusivity\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain exclusivity clause language?\", \"criteria\": {\"true\": \"A exclusivity clause is present in the text.\", \"false\": \"No exclusivity clause language appears.\"}}, \"flag_parties\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain parties clause language?\", \"criteria\": {\"true\": \"A parties clause is present in the text.\", \"false\": \"No parties clause language appears.\"}}, \"flag_effective_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain effective date clause language?\", \"criteria\": {\"true\": \"A effective date clause is present in the text.\", \"false\": \"No effective date clause language appears.\"}}}}}"}], "source": "cuad_packC"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nThis Limited Power of Attorney shall be governed and construed in accordance with the laws of the Commonwealth of Pennsylvania without reference to principles of conflicts of laws."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"clause flag check\", \"request\": {\"model\": \"clef-omni\", \"state\": \"This Limited Power of Attorney shall be governed and construed in accordance with the laws of the Commonwealth of Pennsylvania without reference to principles of conflicts of laws.\", \"questions\": {\"flag_governing_law\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain governing law clause language?\", \"criteria\": {\"true\": \"A governing law clause is present in the text.\", \"false\": \"No governing law clause language appears.\"}}}}}"}], "source": "cuad_packA"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nIf second panel said Socialists, third panel would be something like \"Death panel has determined your life not valuable enough to save\""}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"emotion tagging\", \"request\": {\"model\": \"clef-omni\", \"state\": \"If second panel said Socialists, third panel would be something like \\\"Death panel has determined your life not valuable enough to save\\\"\", \"questions\": {\"emotion_optimism\": {\"type\": \"noul\", \"instructions\": \"Does the author express optimism?\"}}}}"}], "source": "go_emotions"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nLittle known fact. [NAME] died at the age of 8 from small pox."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"emotion tagging\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Little known fact. [NAME] died at the age of 8 from small pox.\", \"questions\": {\"emotion_neutral\": {\"type\": \"noul\", \"instructions\": \"Does the author express neutral?\"}}}}"}], "source": "go_emotions"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nCan I open an account for my offspring?"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Can I open an account for my offspring?\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"pending_transfer\": \"pending transfer\", \"card_swallowed\": \"card swallowed\", \"age_limit\": \"age limit\", \"card_about_to_expire\": \"card about to expire\", \"cash_withdrawal_charge\": \"cash withdrawal charge\", \"exchange_charge\": \"exchange charge\", \"country_support\": \"country support\", \"cancel_transfer\": \"cancel transfer\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nThis Agreement will be in effect for three (3) years from the Effective Date (\"Initial Term\") unless terminated earlier in accordance with this Agreement. ... Zynga shall permit such records to be examined by authorized representatives of WPT, including such independent auditors as WPT may designate, during usual business hours, with advance notice, to verify to the extent necessary the Royalties paid hereunder, and WPT and its representatives shall use reasonable efforts to minimize disruptions to..."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"This Agreement will be in effect for three (3) years from the Effective Date (\\\"Initial Term\\\") unless terminated earlier in accordance with this Agreement. ... Zynga shall permit such records to be examined by authorized representatives of WPT, including such independent auditors as WPT may designate, during usual business hours, with advance notice, to verify to the extent necessary the Royalties paid hereunder, and WPT and its representatives shall use reasonable efforts to minimize disruptions to...\", \"questions\": {\"flag_parties\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain parties clause language?\", \"criteria\": {\"true\": \"A parties clause is present in the text.\", \"false\": \"No parties clause language appears.\"}}, \"flag_most_favored_nation\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain most favored nation clause language?\", \"criteria\": {\"true\": \"A most favored nation clause is present in the text.\", \"false\": \"No most favored nation clause language appears.\"}}, \"flag_non_transferable_license\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain non-transferable license clause language?\", \"criteria\": {\"true\": \"A non-transferable license clause is present in the text.\", \"false\": \"No non-transferable license clause language appears.\"}}, \"flag_cap_on_liability\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain cap on liability clause language?\", \"criteria\": {\"true\": \"A cap on liability clause is present in the text.\", \"false\": \"No cap on liability clause language appears.\"}}, \"flag_minimum_commitment\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain minimum commitment clause language?\", \"criteria\": {\"true\": \"A minimum commitment clause is present in the text.\", \"false\": \"No minimum commitment clause language appears.\"}}, \"flag_ip_ownership_assignment\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain ip ownership assignment clause language?\", \"criteria\": {\"true\": \"A ip ownership assignment clause is present in the text.\", \"false\": \"No ip ownership assignment clause language appears.\"}}, \"flag_audit_rights\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain audit rights clause language?\", \"criteria\": {\"true\": \"A audit rights clause is present in the text.\", \"false\": \"No audit rights clause language appears.\"}}}}}"}], "source": "cuad_packD"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nContract review against the standard 41-clause catalog.\n\nExcerpt:\nSERVICES AGREEMENT ... This Agreement is governed by the laws of England and Wales."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Contract review against the standard 41-clause catalog.\\n\\nExcerpt:\\nSERVICES AGREEMENT ... This Agreement is governed by the laws of England and Wales.\", \"questions\": {\"flag_document_name\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain document name clause language?\", \"criteria\": {\"true\": \"A document name clause is present in the text.\", \"false\": \"No document name clause language appears.\"}}}}}"}], "source": "cuad_packB"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\n@user im just a 19 year old still on loan for the 3rd time at coventry 79 overall hopefully when i return bale will still be ther"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"sentiment rating\", \"request\": {\"model\": \"clef-omni\", \"state\": \"@user im just a 19 year old still on loan for the 3rd time at coventry 79 overall hopefully when i return bale will still be ther\", \"questions\": {\"sentiment\": {\"type\": \"score\", \"instructions\": \"How does the author feel in this message?\", \"criteria\": [\"Negative tone\", \"Neutral, no strong tone\", \"Positive tone\"]}, \"is_complaint\": {\"type\": \"noul\", \"instructions\": \"Is the author complaining about a product or service?\"}}}}"}], "source": "sentiment"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nIf you ever wondered where ISIS got their method of killing homosexuals, wonder no more. Source: al-Jawab al-Kafi p. 260-263 as Quoted by Islamqa:"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"emotion tagging\", \"request\": {\"model\": \"clef-omni\", \"state\": \"If you ever wondered where ISIS got their method of killing homosexuals, wonder no more. Source: al-Jawab al-Kafi p. 260-263 as Quoted by Islamqa:\", \"questions\": {\"emotion_surprise\": {\"type\": \"noul\", \"instructions\": \"Does the author express surprise?\"}, \"emotion_neutral\": {\"type\": \"noul\", \"instructions\": \"Does the author express neutral?\"}}}}"}], "source": "go_emotions"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nMay 29, 2018 ... Sutro shall allocate its available Components and manufacturing capacity to provide SutroVax with quantities of such Product at least equal to the greater of (a) [***] of the amount of Product (or products equivalent to Product) that Sutro allocates for itself and its Affiliates (but in no event less than [***] liters of Extract per month and the..."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"May 29, 2018 ... Sutro shall allocate its available Components and manufacturing capacity to provide SutroVax with quantities of such Product at least equal to the greater of (a) [***] of the amount of Product (or products equivalent to Product) that Sutro allocates for itself and its Affiliates (but in no event less than [***] liters of Extract per month and the...\", \"questions\": {\"flag_post_termination_services\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain post-termination services clause language?\", \"criteria\": {\"true\": \"A post-termination services clause is present in the text.\", \"false\": \"No post-termination services clause language appears.\"}}, \"flag_governing_law\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain governing law clause language?\", \"criteria\": {\"true\": \"A governing law clause is present in the text.\", \"false\": \"No governing law clause language appears.\"}}, \"flag_change_of_control\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain change of control clause language?\", \"criteria\": {\"true\": \"A change of control clause is present in the text.\", \"false\": \"No change of control clause language appears.\"}}, \"flag_parties\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain parties clause language?\", \"criteria\": {\"true\": \"A parties clause is present in the text.\", \"false\": \"No parties clause language appears.\"}}, \"flag_audit_rights\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain audit rights clause language?\", \"criteria\": {\"true\": \"A audit rights clause is present in the text.\", \"false\": \"No audit rights clause language appears.\"}}, \"flag_competitive_restriction_exception\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain competitive restriction exception clause language?\", \"criteria\": {\"true\": \"A competitive restriction exception clause is present in the text.\", \"false\": \"No competitive restriction exception clause language appears.\"}}, \"flag_termination_for_convenience\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain termination for convenience clause language?\", \"criteria\": {\"true\": \"A termination for convenience clause is present in the text.\", \"false\": \"No termination for convenience clause language appears.\"}}, \"flag_insurance\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain insurance clause language?\", \"criteria\": {\"true\": \"A insurance clause is present in the text.\", \"false\": \"No insurance clause language appears.\"}}}}}"}], "source": "cuad_packC"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nZynga shall permit such records to be examined by authorized representatives of WPT, including such independent auditors as WPT may designate, during usual business hours, with advance notice, to verify to the extent necessary the Royalties paid hereunder, and WPT and its representatives shall use reasonable efforts to minimize disruptions to Zynga's business."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"clause flag check\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Zynga shall permit such records to be examined by authorized representatives of WPT, including such independent auditors as WPT may designate, during usual business hours, with advance notice, to verify to the extent necessary the Royalties paid hereunder, and WPT and its representatives shall use reasonable efforts to minimize disruptions to Zynga's business.\", \"questions\": {\"flag_audit_rights\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain audit rights clause language?\", \"criteria\": {\"true\": \"A audit rights clause is present in the text.\", \"false\": \"No audit rights clause language appears.\"}}}}}"}], "source": "cuad_packA"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nThere's no bigger turn off while sexting than someone that doesn't know the difference between 'your' and 'you're'"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"hate speech screen\", \"request\": {\"model\": \"clef-omni\", \"state\": \"There's no bigger turn off while sexting than someone that doesn't know the difference between 'your' and 'you're'\", \"questions\": {\"is_hate_speech\": {\"type\": \"noul\", \"instructions\": \"Does this message contain hate speech against a group?\", \"criteria\": {\"true\": \"Targets a protected group with hate.\", \"false\": \"No hate toward a protected group.\"}}}}}"}], "source": "hate"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nThis Agreement will be in effect for three (3) years from the Effective Date (\"Initial Term\") unless terminated earlier in accordance with this Agreement. ... Zynga shall permit such records to be examined by authorized representatives of WPT, including such independent auditors as WPT may designate, during usual business hours, with advance notice, to verify to the extent necessary the Royalties paid hereunder, and WPT and its representatives shall use reasonable efforts to minimize disruptions to..."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"This Agreement will be in effect for three (3) years from the Effective Date (\\\"Initial Term\\\") unless terminated earlier in accordance with this Agreement. ... Zynga shall permit such records to be examined by authorized representatives of WPT, including such independent auditors as WPT may designate, during usual business hours, with advance notice, to verify to the extent necessary the Royalties paid hereunder, and WPT and its representatives shall use reasonable efforts to minimize disruptions to...\", \"questions\": {\"flag_audit_rights\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain audit rights clause language?\", \"criteria\": {\"true\": \"A audit rights clause is present in the text.\", \"false\": \"No audit rights clause language appears.\"}}, \"flag_parties\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain parties clause language?\", \"criteria\": {\"true\": \"A parties clause is present in the text.\", \"false\": \"No parties clause language appears.\"}}, \"flag_most_favored_nation\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain most favored nation clause language?\", \"criteria\": {\"true\": \"A most favored nation clause is present in the text.\", \"false\": \"No most favored nation clause language appears.\"}}, \"flag_non_transferable_license\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain non-transferable license clause language?\", \"criteria\": {\"true\": \"A non-transferable license clause is present in the text.\", \"false\": \"No non-transferable license clause language appears.\"}}, \"flag_cap_on_liability\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain cap on liability clause language?\", \"criteria\": {\"true\": \"A cap on liability clause is present in the text.\", \"false\": \"No cap on liability clause language appears.\"}}, \"flag_minimum_commitment\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain minimum commitment clause language?\", \"criteria\": {\"true\": \"A minimum commitment clause is present in the text.\", \"false\": \"No minimum commitment clause language appears.\"}}, \"flag_ip_ownership_assignment\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain ip ownership assignment clause language?\", \"criteria\": {\"true\": \"A ip ownership assignment clause is present in the text.\", \"false\": \"No ip ownership assignment clause language appears.\"}}}}}"}], "source": "cuad_packC"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nIs it possible for a friend to personally send me a payment?"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Is it possible for a friend to personally send me a payment?\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"fiat_currency_support\": \"fiat currency support\", \"receiving_money\": \"receiving money\", \"passcode_forgotten\": \"passcode forgotten\", \"transfer_into_account\": \"transfer into account\", \"verify_source_of_funds\": \"verify source of funds\", \"direct_debit_payment_not_recognised\": \"direct debit payment not recognised\", \"card_about_to_expire\": \"card about to expire\", \"pin_blocked\": \"pin blocked\", \"top_up_reverted\": \"top up reverted\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nContract review against the standard 41-clause catalog.\n\nExcerpt:\nLifeUSA/ Envision Health, Inc. ... This Agreement is deemed to have been entered into in the State of Colorado, and its interpretation, construction, and the remedies for its enforcement or breach are to be applied pursuant to and in accordance with the laws of the State of Colorado."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Contract review against the standard 41-clause catalog.\\n\\nExcerpt:\\nLifeUSA/ Envision Health, Inc. ... This Agreement is deemed to have been entered into in the State of Colorado, and its interpretation, construction, and the remedies for its enforcement or breach are to be applied pursuant to and in accordance with the laws of the State of Colorado.\", \"questions\": {\"flag_rofr_rofo_rofn\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain rofr/rofo/rofn clause language?\", \"criteria\": {\"true\": \"A rofr/rofo/rofn clause is present in the text.\", \"false\": \"No rofr/rofo/rofn clause language appears.\"}}}}}"}], "source": "cuad_packB"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nI think the exchange rate is wrong for my ATM withdrawal"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"I think the exchange rate is wrong for my ATM withdrawal\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"cancel_transfer\": \"cancel transfer\", \"pending_top_up\": \"pending top up\", \"receiving_money\": \"receiving money\", \"wrong_exchange_rate_for_cash_withdrawal\": \"wrong exchange rate for cash withdrawal\", \"direct_debit_payment_not_recognised\": \"direct debit payment not recognised\", \"disposable_card_limits\": \"disposable card limits\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nContract review against the standard 41-clause catalog.\n\nExcerpt:\nDuring the Term of this Agreement and for a period of twelve (12) months thereafter, the Distributor (on behalf of itself, each of its affiliates and each of their respective representatives) agrees that it will not directly or indirectly solicit or hire any executive, managerial or technical employee of the Company or any of its affiliates. ... If Distributor complies with all of the terms of this Agreement, the Agreement shall be renewable on an annual basis for one (1) year terms for up to another ten (10) years on the same terms and conditions as set forth herein."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Contract review against the standard 41-clause catalog.\\n\\nExcerpt:\\nDuring the Term of this Agreement and for a period of twelve (12) months thereafter, the Distributor (on behalf of itself, each of its affiliates and each of their respective representatives) agrees that it will not directly or indirectly solicit or hire any executive, managerial or technical employee of the Company or any of its affiliates. ... If Distributor complies with all of the terms of this Agreement, the Agreement shall be renewable on an annual basis for one (1) year terms for up to another ten (10) years on the same terms and conditions as set forth herein.\", \"questions\": {\"flag_effective_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain effective date clause language?\", \"criteria\": {\"true\": \"A effective date clause is present in the text.\", \"false\": \"No effective date clause language appears.\"}}}}}"}], "source": "cuad_packB"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nShopping on line I was declined when trying to buy something."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Shopping on line I was declined when trying to buy something.\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"activate_my_card\": \"activate my card\", \"topping_up_by_card\": \"topping up by card\", \"declined_transfer\": \"declined transfer\", \"apple_pay_or_google_pay\": \"apple pay or google pay\", \"edit_personal_details\": \"edit personal details\", \"card_swallowed\": \"card swallowed\", \"pending_transfer\": \"pending transfer\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation (JSON state):\n{\"subject\": \"Inquiring About Securing Medical Data Using Netgear Nighthawk Mesh Wi-Fi 6 System\", \"message\": \"I need assistance in securing medical data while using the Netgear Nighthawk Mesh Wi-Fi 6 System. Could you provide details on how to ensure the confidentiality and integrity of sensitive medical information that is transmitted through the network? I would greatly appreciate guidance on the necessary configuration settings and additional security measures that can be taken to protect the data.\"}"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"support ticket triage\", \"request\": {\"model\": \"clef-omni\", \"state\": {\"subject\": \"Inquiring About Securing Medical Data Using Netgear Nighthawk Mesh Wi-Fi 6 System\", \"message\": \"I need assistance in securing medical data while using the Netgear Nighthawk Mesh Wi-Fi 6 System. Could you provide details on how to ensure the confidentiality and integrity of sensitive medical information that is transmitted through the network? I would greatly appreciate guidance on the necessary configuration settings and additional security measures that can be taken to protect the data.\"}, \"questions\": {\"route_to\": {\"type\": \"choice\", \"instructions\": \"Which support queue should handle this ticket?\", \"criteria\": {\"IT & Technology/Network Infrastructure\": \"IT & Technology/Network Infrastructure\", \"IT Support\": \"IT Support\", \"Returns and Exchanges\": \"Returns and Exchanges\", \"General Inquiry\": \"General Inquiry\", \"Human Resources\": \"Human Resources\"}}, \"ticket_type\": {\"type\": \"choice\", \"instructions\": \"What kind of ticket is this?\", \"criteria\": {\"Change\": \"Change\", \"Incident\": \"Incident\", \"Problem\": \"Problem\", \"Request\": \"Request\"}}, \"priority\": {\"type\": \"score\", \"instructions\": \"How urgent is this ticket?\", \"criteria\": [\"Low, can wait days\", \"Medium, handle this week\", \"High, handle today\", \"Critical, service blocked\"]}, \"needs_escalation\": {\"type\": \"noul\", \"instructions\": \"Should this be escalated beyond first-line support?\"}}}}"}], "source": "tickets"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nTold my dad I was going to wear all black to my aunts house & he just said it's not Halloween yet Wednesday Adams lol"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"sentiment rating\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Told my dad I was going to wear all black to my aunts house & he just said it's not Halloween yet Wednesday Adams lol\", \"questions\": {\"sentiment\": {\"type\": \"score\", \"instructions\": \"How does the author feel in this message?\", \"criteria\": [\"Negative tone\", \"Neutral, no strong tone\", \"Positive tone\"]}, \"is_complaint\": {\"type\": \"noul\", \"instructions\": \"Is the author complaining about a product or service?\"}}}}"}], "source": "sentiment"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nSERVICES AGREEMENT ... This Agreement is governed by the laws of England and Wales."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"SERVICES AGREEMENT ... This Agreement is governed by the laws of England and Wales.\", \"questions\": {\"flag_document_name\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain document name clause language?\", \"criteria\": {\"true\": \"A document name clause is present in the text.\", \"false\": \"No document name clause language appears.\"}}, \"flag_anti_assignment\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain anti-assignment clause language?\", \"criteria\": {\"true\": \"A anti-assignment clause is present in the text.\", \"false\": \"No anti-assignment clause language appears.\"}}, \"flag_no_solicit_of_employees\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain no-solicit of employees clause language?\", \"criteria\": {\"true\": \"A no-solicit of employees clause is present in the text.\", \"false\": \"No no-solicit of employees clause language appears.\"}}, \"flag_termination_for_convenience\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain termination for convenience clause language?\", \"criteria\": {\"true\": \"A termination for convenience clause is present in the text.\", \"false\": \"No termination for convenience clause language appears.\"}}, \"flag_effective_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain effective date clause language?\", \"criteria\": {\"true\": \"A effective date clause is present in the text.\", \"false\": \"No effective date clause language appears.\"}}, \"flag_expiration_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain expiration date clause language?\", \"criteria\": {\"true\": \"A expiration date clause is present in the text.\", \"false\": \"No expiration date clause language appears.\"}}, \"flag_parties\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain parties clause language?\", \"criteria\": {\"true\": \"A parties clause is present in the text.\", \"false\": \"No parties clause language appears.\"}}}}}"}], "source": "cuad_packC"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nWho gets support?"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Who gets support?\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"wrong_exchange_rate_for_cash_withdrawal\": \"wrong exchange rate for cash withdrawal\", \"supported_cards_and_currencies\": \"supported cards and currencies\", \"verify_my_identity\": \"verify my identity\", \"transfer_timing\": \"transfer timing\", \"country_support\": \"country support\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\n\"Frank Ocean, you better be working on that album Saturday night since you cancelled your FYF appearance.\""}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"sentiment rating\", \"request\": {\"model\": \"clef-omni\", \"state\": \"\\\"Frank Ocean, you better be working on that album Saturday night since you cancelled your FYF appearance.\\\"\", \"questions\": {\"sentiment\": {\"type\": \"score\", \"instructions\": \"How does the author feel in this message?\", \"criteria\": [\"Negative tone\", \"Neutral, no strong tone\", \"Positive tone\"]}, \"is_complaint\": {\"type\": \"noul\", \"instructions\": \"Is the author complaining about a product or service?\"}}}}"}], "source": "sentiment"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nTop up is not working even though I have my AMEX in apple pay."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Top up is not working even though I have my AMEX in apple pay.\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"apple_pay_or_google_pay\": \"apple pay or google pay\", \"age_limit\": \"age limit\", \"card_payment_wrong_exchange_rate\": \"card payment wrong exchange rate\", \"request_refund\": \"request refund\", \"card_not_working\": \"card not working\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation (JSON state):\n{\"subject\": \"Problem with Investment Reports Generation\", \"message\": \"Investment optimization reports are not generating correctly after recent software updates. Restarting the applications and clearing caches did not resolve the issue.\"}"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"support ticket triage\", \"request\": {\"model\": \"clef-omni\", \"state\": {\"subject\": \"Problem with Investment Reports Generation\", \"message\": \"Investment optimization reports are not generating correctly after recent software updates. Restarting the applications and clearing caches did not resolve the issue.\"}, \"questions\": {\"route_to\": {\"type\": \"choice\", \"instructions\": \"Which support queue should handle this ticket?\", \"criteria\": {\"People & Society/Culture & Society\": \"People & Society/Culture & Society\", \"IT & Technology/Network Infrastructure\": \"IT & Technology/Network Infrastructure\", \"Technical Support\": \"Technical Support\", \"Service Outages and Maintenance\": \"Service Outages and Maintenance\", \"Food & Drink/Restaurants\": \"Food & Drink/Restaurants\"}}, \"ticket_type\": {\"type\": \"choice\", \"instructions\": \"What kind of ticket is this?\", \"criteria\": {\"Request\": \"Request\", \"Incident\": \"Incident\", \"Change\": \"Change\", \"Problem\": \"Problem\"}}, \"priority\": {\"type\": \"score\", \"instructions\": \"How urgent is this ticket?\", \"criteria\": [\"Low, can wait days\", \"Medium, handle this week\", \"High, handle today\", \"Critical, service blocked\"]}, \"needs_escalation\": {\"type\": \"noul\", \"instructions\": \"Should this be escalated beyond first-line support?\"}}}}"}], "source": "tickets"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nI had a payment reverted and would like to figure out why."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"I had a payment reverted and would like to figure out why.\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"balance_not_updated_after_bank_transfer\": \"balance not updated after bank transfer\", \"change_pin\": \"change pin\", \"reverted_card_payment?\": \"reverted card payment?\", \"cancel_transfer\": \"cancel transfer\", \"country_support\": \"country support\", \"card_payment_wrong_exchange_rate\": \"card payment wrong exchange rate\", \"cash_withdrawal_not_recognised\": \"cash withdrawal not recognised\", \"passcode_forgotten\": \"passcode forgotten\", \"failed_transfer\": \"failed transfer\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation (JSON state):\n{\"subject\": \"Support for Digital Marketing with Mailchimp\", \"message\": \"Is it possible to get details on integrating Mailchimp digital marketing strategies? We aim to enhance our email campaigns and automate our workflow. Thank you for your assistance, and I look forward to hearing back soon.\"}"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"support ticket triage\", \"request\": {\"model\": \"clef-omni\", \"state\": {\"subject\": \"Support for Digital Marketing with Mailchimp\", \"message\": \"Is it possible to get details on integrating Mailchimp digital marketing strategies? We aim to enhance our email campaigns and automate our workflow. Thank you for your assistance, and I look forward to hearing back soon.\"}, \"questions\": {\"route_to\": {\"type\": \"choice\", \"instructions\": \"Which support queue should handle this ticket?\", \"criteria\": {\"Service Outages and Maintenance\": \"Service Outages and Maintenance\", \"General Inquiry\": \"General Inquiry\", \"IT Support\": \"IT Support\", \"Law & Government/Legal Advice\": \"Law & Government/Legal Advice\", \"Pets & Animals/Pet Services\": \"Pets & Animals/Pet Services\"}}, \"ticket_type\": {\"type\": \"choice\", \"instructions\": \"What kind of ticket is this?\", \"criteria\": {\"Change\": \"Change\", \"Problem\": \"Problem\", \"Request\": \"Request\", \"Incident\": \"Incident\"}}, \"priority\": {\"type\": \"score\", \"instructions\": \"How urgent is this ticket?\", \"criteria\": [\"Low, can wait days\", \"Medium, handle this week\", \"High, handle today\", \"Critical, service blocked\"]}, \"needs_escalation\": {\"type\": \"noul\", \"instructions\": \"Should this be escalated beyond first-line support?\"}}}}"}], "source": "tickets"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nUnsure how to provide my identity."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Unsure how to provide my identity.\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"visa_or_mastercard\": \"visa or mastercard\", \"card_not_working\": \"card not working\", \"get_disposable_virtual_card\": \"get disposable virtual card\", \"unable_to_verify_identity\": \"unable to verify identity\", \"transfer_fee_charged\": \"transfer fee charged\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nWhy hasn't my top-up gone through?"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Why hasn't my top-up gone through?\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"exchange_via_app\": \"exchange via app\", \"failed_transfer\": \"failed transfer\", \"pending_transfer\": \"pending transfer\", \"card_payment_fee_charged\": \"card payment fee charged\", \"reverted_card_payment?\": \"reverted card payment?\", \"unable_to_verify_identity\": \"unable to verify identity\", \"pending_top_up\": \"pending top up\", \"lost_or_stolen_card\": \"lost or stolen card\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nContract review against the standard 41-clause catalog.\n\nExcerpt:\nDecember 14, 2015 ... TERM: November 01, 2014 to October 31, 2017"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Contract review against the standard 41-clause catalog.\\n\\nExcerpt:\\nDecember 14, 2015 ... TERM: November 01, 2014 to October 31, 2017\", \"questions\": {\"flag_agreement_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain agreement date clause language?\", \"criteria\": {\"true\": \"A agreement date clause is present in the text.\", \"false\": \"No agreement date clause language appears.\"}}}}}"}], "source": "cuad_packB"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nI’ve made candies almonds and cashews in a slow cooker before. They’re surprisingly not mushy or soggy."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"emotion tagging\", \"request\": {\"model\": \"clef-omni\", \"state\": \"I’ve made candies almonds and cashews in a slow cooker before. They’re surprisingly not mushy or soggy.\", \"questions\": {\"emotion_surprise\": {\"type\": \"noul\", \"instructions\": \"Does the author express surprise?\"}}}}"}], "source": "go_emotions"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nSERVICES AGREEMENT ... This Agreement is governed by the laws of England and Wales."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"SERVICES AGREEMENT ... This Agreement is governed by the laws of England and Wales.\", \"questions\": {\"flag_anti_assignment\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain anti-assignment clause language?\", \"criteria\": {\"true\": \"A anti-assignment clause is present in the text.\", \"false\": \"No anti-assignment clause language appears.\"}}, \"flag_no_solicit_of_employees\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain no-solicit of employees clause language?\", \"criteria\": {\"true\": \"A no-solicit of employees clause is present in the text.\", \"false\": \"No no-solicit of employees clause language appears.\"}}, \"flag_termination_for_convenience\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain termination for convenience clause language?\", \"criteria\": {\"true\": \"A termination for convenience clause is present in the text.\", \"false\": \"No termination for convenience clause language appears.\"}}, \"flag_effective_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain effective date clause language?\", \"criteria\": {\"true\": \"A effective date clause is present in the text.\", \"false\": \"No effective date clause language appears.\"}}, \"flag_expiration_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain expiration date clause language?\", \"criteria\": {\"true\": \"A expiration date clause is present in the text.\", \"false\": \"No expiration date clause language appears.\"}}, \"flag_parties\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain parties clause language?\", \"criteria\": {\"true\": \"A parties clause is present in the text.\", \"false\": \"No parties clause language appears.\"}}, \"flag_document_name\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain document name clause language?\", \"criteria\": {\"true\": \"A document name clause is present in the text.\", \"false\": \"No document name clause language appears.\"}}}}}"}], "source": "cuad_packD"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nNotwithstanding anything to the contrary herein, in the event Company incurs any expenses, damages or other liabilities (including, without limitation, reasonable attorneys' fees) in connection with the performance or non-performance of any term or provision hereof, Licensor's liability to Company shall not exceed the remuneration, excluding reimbursement of expenses, actually paid to Licensor by Company."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"clause flag check\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Notwithstanding anything to the contrary herein, in the event Company incurs any expenses, damages or other liabilities (including, without limitation, reasonable attorneys' fees) in connection with the performance or non-performance of any term or provision hereof, Licensor's liability to Company shall not exceed the remuneration, excluding reimbursement of expenses, actually paid to Licensor by Company.\", \"questions\": {\"flag_cap_on_liability\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain cap on liability clause language?\", \"criteria\": {\"true\": \"A cap on liability clause is present in the text.\", \"false\": \"No cap on liability clause language appears.\"}}}}}"}], "source": "cuad_packA"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nSorry it didn't help. I hope you can find something that does."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"emotion tagging\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Sorry it didn't help. I hope you can find something that does.\", \"questions\": {\"emotion_optimism\": {\"type\": \"noul\", \"instructions\": \"Does the author express optimism?\"}, \"emotion_remorse\": {\"type\": \"noul\", \"instructions\": \"Does the author express remorse?\"}}}}"}], "source": "go_emotions"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation (JSON state):\n{\"subject\": \"Inquiry on System Requirements for SaaS Project Management Tool\", \"message\": \"Hello Customer Support, I am writing to request information on the system requirements needed for the best performance of your SaaS project management tool. Could you kindly share the recommended browser, operating system, and hardware specifications? Additionally, any extra details on how to achieve the best user experience would be greatly appreciated. I look forward to your response and thank you in advance for your help.\"}"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"support ticket triage\", \"request\": {\"model\": \"clef-omni\", \"state\": {\"subject\": \"Inquiry on System Requirements for SaaS Project Management Tool\", \"message\": \"Hello Customer Support, I am writing to request information on the system requirements needed for the best performance of your SaaS project management tool. Could you kindly share the recommended browser, operating system, and hardware specifications? Additionally, any extra details on how to achieve the best user experience would be greatly appreciated. I look forward to your response and thank you in advance for your help.\"}, \"questions\": {\"route_to\": {\"type\": \"choice\", \"instructions\": \"Which support queue should handle this ticket?\", \"criteria\": {\"IT & Technology/Software Development\": \"IT & Technology/Software Development\", \"Science/Environmental Science\": \"Science/Environmental Science\", \"Technical Support\": \"Technical Support\", \"Travel & Transportation/Land Travel\": \"Travel & Transportation/Land Travel\", \"Food & Drink/Groceries\": \"Food & Drink/Groceries\"}}, \"ticket_type\": {\"type\": \"choice\", \"instructions\": \"What kind of ticket is this?\", \"criteria\": {\"Request\": \"Request\", \"Problem\": \"Problem\", \"Change\": \"Change\", \"Incident\": \"Incident\"}}, \"priority\": {\"type\": \"score\", \"instructions\": \"How urgent is this ticket?\", \"criteria\": [\"Low, can wait days\", \"Medium, handle this week\", \"High, handle today\", \"Critical, service blocked\"]}, \"needs_escalation\": {\"type\": \"noul\", \"instructions\": \"Should this be escalated beyond first-line support?\"}}}}"}], "source": "tickets"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\n\"Wisconsin Gov. Scott Walker talked tough on immigration Monday, saying he had a \"\"similar\"\" approach to that of real estate mogul Donald\""}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"sentiment rating\", \"request\": {\"model\": \"clef-omni\", \"state\": \"\\\"Wisconsin Gov. Scott Walker talked tough on immigration Monday, saying he had a \\\"\\\"similar\\\"\\\" approach to that of real estate mogul Donald\\\"\", \"questions\": {\"sentiment\": {\"type\": \"score\", \"instructions\": \"How does the author feel in this message?\", \"criteria\": [\"Negative tone\", \"Neutral, no strong tone\", \"Positive tone\"]}, \"is_complaint\": {\"type\": \"noul\", \"instructions\": \"Is the author complaining about a product or service?\"}}}}"}], "source": "sentiment"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\n@user You may want to learn about what the SCOTUS does. @user"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"sentiment rating\", \"request\": {\"model\": \"clef-omni\", \"state\": \"@user You may want to learn about what the SCOTUS does. @user\", \"questions\": {\"sentiment\": {\"type\": \"score\", \"instructions\": \"How does the author feel in this message?\", \"criteria\": [\"Negative tone\", \"Neutral, no strong tone\", \"Positive tone\"]}, \"is_complaint\": {\"type\": \"noul\", \"instructions\": \"Is the author complaining about a product or service?\"}}}}"}], "source": "sentiment"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation (JSON state):\n{\"subject\": \"Problem with Login SaaS Platform\", \"message\": \"Experienced occasional login difficulties when accessing the SaaS platform, which might be related to backend server overload. Despite clearing server caches and restarting services, the issue remains. Need your help to address this.\"}"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"support ticket triage\", \"request\": {\"model\": \"clef-omni\", \"state\": {\"subject\": \"Problem with Login SaaS Platform\", \"message\": \"Experienced occasional login difficulties when accessing the SaaS platform, which might be related to backend server overload. Despite clearing server caches and restarting services, the issue remains. Need your help to address this.\"}, \"questions\": {\"route_to\": {\"type\": \"choice\", \"instructions\": \"Which support queue should handle this ticket?\", \"criteria\": {\"Product Support\": \"Product Support\", \"IT Support\": \"IT Support\", \"Business & Industrial/Manufacturing\": \"Business & Industrial/Manufacturing\", \"IT & Technology/Security Operations\": \"IT & Technology/Security Operations\", \"Arts & Entertainment/Music\": \"Arts & Entertainment/Music\"}}, \"ticket_type\": {\"type\": \"choice\", \"instructions\": \"What kind of ticket is this?\", \"criteria\": {\"Incident\": \"Incident\", \"Change\": \"Change\", \"Problem\": \"Problem\", \"Request\": \"Request\"}}, \"priority\": {\"type\": \"score\", \"instructions\": \"How urgent is this ticket?\", \"criteria\": [\"Low, can wait days\", \"Medium, handle this week\", \"High, handle today\", \"Critical, service blocked\"]}, \"needs_escalation\": {\"type\": \"noul\", \"instructions\": \"Should this be escalated beyond first-line support?\"}}}}"}], "source": "tickets"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation (JSON state):\n{\"subject\": \"Assistance with Keras RAID Setup\", \"message\": \"Customer Support, I am seeking comprehensive instructions for configuring the Keras RAID controller within financial data analytics platforms. The objective is to enhance the performance and reliability of our data processing workflows. Can you provide a detailed, step-by-step procedure to set up the RAID controller for smooth integration with Keras? Additionally, any specific hardware and software prerequisites would be beneficial. We would appreciate any resources or guidance to facilitate a successful implementation.\"}"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"support ticket triage\", \"request\": {\"model\": \"clef-omni\", \"state\": {\"subject\": \"Assistance with Keras RAID Setup\", \"message\": \"Customer Support, I am seeking comprehensive instructions for configuring the Keras RAID controller within financial data analytics platforms. The objective is to enhance the performance and reliability of our data processing workflows. Can you provide a detailed, step-by-step procedure to set up the RAID controller for smooth integration with Keras? Additionally, any specific hardware and software prerequisites would be beneficial. We would appreciate any resources or guidance to facilitate a successful implementation.\"}, \"questions\": {\"route_to\": {\"type\": \"choice\", \"instructions\": \"Which support queue should handle this ticket?\", \"criteria\": {\"Service Outages and Maintenance\": \"Service Outages and Maintenance\", \"IT & Technology/Network Infrastructure\": \"IT & Technology/Network Infrastructure\", \"Law & Government/Government Services\": \"Law & Government/Government Services\", \"Home & Garden/Home Improvement\": \"Home & Garden/Home Improvement\", \"Jobs & Education/Online Courses\": \"Jobs & Education/Online Courses\"}}, \"ticket_type\": {\"type\": \"choice\", \"instructions\": \"What kind of ticket is this?\", \"criteria\": {\"Problem\": \"Problem\", \"Change\": \"Change\", \"Incident\": \"Incident\", \"Request\": \"Request\"}}, \"priority\": {\"type\": \"score\", \"instructions\": \"How urgent is this ticket?\", \"criteria\": [\"Low, can wait days\", \"Medium, handle this week\", \"High, handle today\", \"Critical, service blocked\"]}, \"needs_escalation\": {\"type\": \"noul\", \"instructions\": \"Should this be escalated beyond first-line support?\"}}}}"}], "source": "tickets"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nit was amazing"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"emotion tagging\", \"request\": {\"model\": \"clef-omni\", \"state\": \"it was amazing\", \"questions\": {\"emotion_admiration\": {\"type\": \"noul\", \"instructions\": \"Does the author express admiration?\"}}}}"}], "source": "go_emotions"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nMy statement has a payment that I do not recall making"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"banking intent routing\", \"request\": {\"model\": \"clef-omni\", \"state\": \"My statement has a payment that I do not recall making\", \"questions\": {\"intent\": {\"type\": \"choice\", \"instructions\": \"What does the customer want? Which banking intent fits this message?\", \"criteria\": {\"card_arrival\": \"card arrival\", \"transfer_timing\": \"transfer timing\", \"order_physical_card\": \"order physical card\", \"lost_or_stolen_phone\": \"lost or stolen phone\", \"card_linking\": \"card linking\", \"card_payment_not_recognised\": \"card payment not recognised\", \"cash_withdrawal_not_recognised\": \"cash withdrawal not recognised\"}}}}}"}], "source": "banking77"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation (JSON state):\n{\"subject\": \"Analysis of Issue Tracking Metrics\", \"message\": \"The tracking analytics indicate precise campaign performance metrics following the recent software updates.\"}"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"support ticket triage\", \"request\": {\"model\": \"clef-omni\", \"state\": {\"subject\": \"Analysis of Issue Tracking Metrics\", \"message\": \"The tracking analytics indicate precise campaign performance metrics following the recent software updates.\"}, \"questions\": {\"route_to\": {\"type\": \"choice\", \"instructions\": \"Which support queue should handle this ticket?\", \"criteria\": {\"General Inquiry\": \"General Inquiry\", \"Sales and Pre-Sales\": \"Sales and Pre-Sales\", \"Games\": \"Games\", \"IT & Technology/Network Infrastructure\": \"IT & Technology/Network Infrastructure\", \"Technical Support\": \"Technical Support\"}}, \"ticket_type\": {\"type\": \"choice\", \"instructions\": \"What kind of ticket is this?\", \"criteria\": {\"Request\": \"Request\", \"Change\": \"Change\", \"Problem\": \"Problem\", \"Incident\": \"Incident\"}}, \"priority\": {\"type\": \"score\", \"instructions\": \"How urgent is this ticket?\", \"criteria\": [\"Low, can wait days\", \"Medium, handle this week\", \"High, handle today\", \"Critical, service blocked\"]}, \"needs_escalation\": {\"type\": \"noul\", \"instructions\": \"Should this be escalated beyond first-line support?\"}}}}"}], "source": "tickets"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nIs that good or bad?"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"emotion tagging\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Is that good or bad?\", \"questions\": {\"emotion_confusion\": {\"type\": \"noul\", \"instructions\": \"Does the author express confusion?\"}}}}"}], "source": "go_emotions"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nContract review against the standard 41-clause catalog.\n\nExcerpt:\nMay 29, 2018 ... Sutro shall allocate its available Components and manufacturing capacity to provide SutroVax with quantities of such Product at least equal to the greater of (a) [***] of the amount of Product (or products equivalent to Product) that Sutro allocates for itself and its Affiliates (but in no event less than [***] liters of Extract per month and the..."}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Contract review against the standard 41-clause catalog.\\n\\nExcerpt:\\nMay 29, 2018 ... Sutro shall allocate its available Components and manufacturing capacity to provide SutroVax with quantities of such Product at least equal to the greater of (a) [***] of the amount of Product (or products equivalent to Product) that Sutro allocates for itself and its Affiliates (but in no event less than [***] liters of Extract per month and the...\", \"questions\": {\"flag_post_termination_services\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain post-termination services clause language?\", \"criteria\": {\"true\": \"A post-termination services clause is present in the text.\", \"false\": \"No post-termination services clause language appears.\"}}}}}"}], "source": "cuad_packB"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nDude, wtf is this?"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"emotion tagging\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Dude, wtf is this?\", \"questions\": {\"emotion_anger\": {\"type\": \"noul\", \"instructions\": \"Does the author express anger?\"}}}}"}], "source": "go_emotions"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation (JSON state):\n{\"subject\": \"Review of Recent Marketing Campaign Outcomes\", \"message\": \"The recent digital marketing campaign did not achieve the anticipated growth for the brand. This might be due to inadequate audience targeting. Initial efforts to analyze the campaign's performance and adjust strategies have shown some improvements. I would greatly appreciate more detailed guidance on how to better target the audience and enhance the campaign's performance.\"}"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"support ticket triage\", \"request\": {\"model\": \"clef-omni\", \"state\": {\"subject\": \"Review of Recent Marketing Campaign Outcomes\", \"message\": \"The recent digital marketing campaign did not achieve the anticipated growth for the brand. This might be due to inadequate audience targeting. Initial efforts to analyze the campaign's performance and adjust strategies have shown some improvements. I would greatly appreciate more detailed guidance on how to better target the audience and enhance the campaign's performance.\"}, \"questions\": {\"route_to\": {\"type\": \"choice\", \"instructions\": \"Which support queue should handle this ticket?\", \"criteria\": {\"Books & Literature/Fiction\": \"Books & Literature/Fiction\", \"Shopping/Retail Stores\": \"Shopping/Retail Stores\", \"Returns and Exchanges\": \"Returns and Exchanges\", \"Jobs & Education/Online Courses\": \"Jobs & Education/Online Courses\", \"Home & Garden/Home Improvement\": \"Home & Garden/Home Improvement\"}}, \"ticket_type\": {\"type\": \"choice\", \"instructions\": \"What kind of ticket is this?\", \"criteria\": {\"Incident\": \"Incident\", \"Request\": \"Request\", \"Change\": \"Change\", \"Problem\": \"Problem\"}}, \"priority\": {\"type\": \"score\", \"instructions\": \"How urgent is this ticket?\", \"criteria\": [\"Low, can wait days\", \"Medium, handle this week\", \"High, handle today\", \"Critical, service blocked\"]}, \"needs_escalation\": {\"type\": \"noul\", \"instructions\": \"Should this be escalated beyond first-line support?\"}}}}"}], "source": "tickets"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nDecember 14, 2015 ... TERM: November 01, 2014 to October 31, 2017"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"contract clause review\", \"request\": {\"model\": \"clef-omni\", \"state\": \"December 14, 2015 ... TERM: November 01, 2014 to October 31, 2017\", \"questions\": {\"flag_agreement_date\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain agreement date clause language?\", \"criteria\": {\"true\": \"A agreement date clause is present in the text.\", \"false\": \"No agreement date clause language appears.\"}}, \"flag_governing_law\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain governing law clause language?\", \"criteria\": {\"true\": \"A governing law clause is present in the text.\", \"false\": \"No governing law clause language appears.\"}}, \"flag_parties\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain parties clause language?\", \"criteria\": {\"true\": \"A parties clause is present in the text.\", \"false\": \"No parties clause language appears.\"}}, \"flag_insurance\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain insurance clause language?\", \"criteria\": {\"true\": \"A insurance clause is present in the text.\", \"false\": \"No insurance clause language appears.\"}}, \"flag_non_transferable_license\": {\"type\": \"noul\", \"instructions\": \"Does this contract excerpt contain non-transferable license clause language?\", \"criteria\": {\"true\": \"A non-transferable license clause is present in the text.\", \"false\": \"No non-transferable license clause language appears.\"}}}}}"}], "source": "cuad_packC"} | |
| {"prompt": [{"role": "system", "content": "You are a Clef-Omni orchestrator. For every user situation you decide how to interrogate Clef-Omni, a decision model served at POST /v1/systemone. Clef-Omni judges a state against a schema of typed questions in a single forward pass and returns a probability for every allowed option of every question.\n\nQuestion types:\n- \"noul\": yes/no. Returns the probability of true. Use when one property, true or false, decides something. Optional criteria: {\"true\": \"...\", \"false\": \"...\"}.\n- \"choice\": exactly one of several labels fits. criteria is an object mapping option id to a short, concrete description. Use 2 to 10 options; descriptions matter more than ids.\n- \"score\": ordered levels from lowest to highest. criteria is a list of 2 to 6 level descriptions. Prefer score over choice when the levels have an order.\n\nRules:\n- 1 to 8 questions per request. All questions are judged against the same state.\n- Several labels can hold at once -> ask one noul question per label in the same request.\n- Question ids: lowercase snake_case, letters, digits, '_', '-', '.', max 100 chars.\n- The state may be a raw string or a JSON object. Pass through the user's situation, trimmed of noise.\n- When several labels can hold at once, or one property needs a yes/no, an ordered scale, and a routing label together, combine the question types in one request.\n- Respond with ONLY a JSON object, no markdown fences:\n{\"use_case\": \"<2-6 words naming the decision>\", \"request\": {\"model\": \"clef-omni\", \"state\": <state>, \"questions\": {<question id>: {\"type\": \"noul\"|\"choice\"|\"score\", \"instructions\": \"...\", \"criteria\": ...}, ...}}}"}, {"role": "user", "content": "Situation:\nReading this reminds me of the witch hunt scene in Holy Grail. Thanks Daily Beast!"}], "completion": [{"role": "assistant", "content": "{\"use_case\": \"emotion tagging\", \"request\": {\"model\": \"clef-omni\", \"state\": \"Reading this reminds me of the witch hunt scene in Holy Grail. Thanks Daily Beast!\", \"questions\": {\"emotion_gratitude\": {\"type\": \"noul\", \"instructions\": \"Does the author express gratitude?\"}}}}"}], "source": "go_emotions"} | |