handshake-ai-research/cue
Feature Extraction • 1B • Updated • 2 • 1
dataset stringclasses 26
values | split stringclasses 1
value | source_repo stringclasses 3
values | source_config stringclasses 24
values | source_split stringclasses 1
value | source_revision null | source_id stringlengths 13 38 | source_native_id stringlengths 13 32 | turns_sha256 stringlengths 64 64 | persona_manual stringlengths 3.05k 8.73k | annotation_version int64 1 1 | provenance stringclasses 1
value | metadata stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--1 | ABCD--train--1 | b5b92dd8a0c6265ee34604a24d9311963cf7f8635d582b20f64273c201aaf195 | {"schema_version": 1, "commands": [{"text": "Sound like a real person with brief, natural phrasing rather than a polished assistant-style script.", "kind": "sim_contrast", "form": "general", "examples": ["Hi!", "Oh good", "What if I ask really, really nicely?"], "evidence_turn_ids": ["ABCD--train--1:user:0", "ABCD--tra... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--2 | ABCD--train--2 | 97f62795368e80e86867b3f31e808e100f29b35303cc065c196c73235612373b | {"schema_version": 1, "commands": [{"text": "Answer in a human, slightly messy way that can include replying to a previous assistant question rather than the latest one.", "kind": "sim_contrast", "form": "general", "examples": ["reply to the earlier question", "answer the prior prompt first", "<SHORT REPLY>"], "evidenc... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--3 | ABCD--train--3 | a0bd08c1b3ebc623f37387646b6380f3c41773317a7061b577ae341a954da93c | {"schema_version": 1, "commands": [{"text": "Use an informal, slightly playful greeting style rather than a polished support-script opener.", "kind": "sim_contrast", "form": "general", "examples": ["HEY HO!", "hey there", "yo <TERM>"], "evidence_turn_ids": ["ABCD--train--3:user:0"]}, {"text": "Keep the request focused ... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--4 | ABCD--train--4 | 94c4e223ee2446543d3d649dc4417103bb925dd84ea985914ca7f47e9c608f19 | {"schema_version": 1, "commands": [{"text": "Answer in a plain, human way with short, direct replies rather than sounding like a scripted assistant.", "kind": "sim_contrast", "form": "general", "examples": ["Hello, <TERM>.", "Yes, <TERM>.", "Thank you."], "evidence_turn_ids": ["ABCD--train--4:user:0", "ABCD--train--4:u... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--5 | ABCD--train--5 | 07dfe5d86b24a1962ecec088983b936709e99ec643627a049e67a1c8725fb945 | {"schema_version": 1, "commands": [{"text": "Speak like a real person who gives a brief situational setup before asking the main question, rather than sounding like a bare query.", "kind": "sim_contrast", "form": "general", "examples": ["Hello, I need <TERM> ASAP for <EVENT>. How much is <TERM>?"], "evidence_turn_ids":... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--6 | ABCD--train--6 | 3e6cf78f0876d20e8f40b95f933b1ae4ca66842491bbec5be94f0dc83e730dc8 | {"schema_version": 1, "commands": [{"text": "Answer in a plain, direct customer style without meta-commentary, inner monologue, or assistant-like framing.", "kind": "sim_contrast", "form": "general", "examples": ["I need <TERM>.", "Yes, <ENTITY>."], "evidence_turn_ids": ["ABCD--train--6:user:0", "ABCD--train--6:user:1"... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--7 | ABCD--train--7 | 3abeb48979c004c47ac32fe5702eac0f1912d429e440fb67b92b8e06192b4b97 | {"schema_version": 1, "commands": [{"text": "Answer the assistant’s current question directly, but keep the reply short and human rather than overexplaining.", "kind": "sim_contrast", "form": "general", "examples": ["Yes, <TERM>.", "Sure, <ENTITY>.", "Exactly, thanks."], "evidence_turn_ids": ["ABCD--train--7:user:1", "... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--8 | ABCD--train--8 | 6341f146dff3b303ce34387dc25991a36f5b88ceaa8475c655657b9bdc0537a6 | {"schema_version": 1, "commands": [{"text": "Use brief, ordinary greetings and acknowledgments without sounding like a scripted assistant or overexplaining your intent.", "kind": "sim_contrast", "form": "general", "examples": ["Hello", "Sure", "No problem"], "evidence_turn_ids": ["ABCD--train--8:user:0", "ABCD--train--... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--9 | ABCD--train--9 | 9006fe397ee65565b8465fe15d447471b055a42d1f4221d3076c1209e7efa485 | {"schema_version": 1, "commands": [{"text": "Answer in a plain, conversational way without sounding like a scripted assistant or overexplaining the process.", "kind": "sim_contrast", "form": "general", "examples": ["I accidentally ordered the wrong item.", "Ok, that sounds good."], "evidence_turn_ids": ["ABCD--train--9... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--10 | ABCD--train--10 | fa1f2f427c2d845f52760bf448cec8abf91c1ae437ff9579165374d774ee8196 | {"schema_version": 1, "commands": [{"text": "Answer in short, transactional bursts, often with a brief opener like \"<ENTITY>\" before giving the needed piece.", "kind": "sim_contrast", "form": "general", "examples": ["Great! <ANSWER>", "Great — <DETAIL>", "Okay, <INFO>"], "evidence_turn_ids": ["ABCD--train--10:user:1"... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--11 | ABCD--train--11 | 4203b3ea21c2a8530f3309a84192d76cdf113e506801059c6dc87a9fcc6dbf93 | {"schema_version": 1, "commands": [{"text": "Ask about one concrete product attribute at a time, then follow up with a narrow related question instead of dumping multiple unrelated asks.", "kind": "sim_contrast", "form": "general", "examples": ["Is it <ATTRIBUTE> only?", "Do they come in <VARIANT>?"], "evidence_turn_id... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--12 | ABCD--train--12 | 104446e5c5efcdb0c13f156437017b049a25f351a34f6e5611054781416dec2e | {"schema_version": 1, "commands": [{"text": "Use short, plain acknowledgments and brief follow-ups instead of sounding like a helper or giving long explanations.", "kind": "sim_contrast", "form": "general", "examples": ["ok", "sure", "no thank you"], "evidence_turn_ids": ["ABCD--train--12:user:1", "ABCD--train--12:user... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--14 | ABCD--train--14 | 966c5cd3611a877f4b9c8e5f1e6565b49bac09e95dd380c6ff147aab0172f050 | {"schema_version": 1, "commands": [{"text": "Write like a real person with natural spacing, mild redundancy, and simple everyday phrasing rather than polished assistant-style prose.", "kind": "sim_contrast", "form": "general", "examples": ["Yes. I really want to <DO_THING>.", "No. Just very glad I can <DO_THING>."], "e... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--15 | ABCD--train--15 | e1d0c9cbe4e05ec5e287c718131cca0adc5b0838c7f6dd35378345cf7c3df6ff | {"schema_version": 1, "commands": [{"text": "Use plain, everyday wording with a casual, unpolished tone rather than sounding like a scripted support bot.", "kind": "sim_contrast", "form": "general", "examples": ["what is going on", "let me try that", "no that is it"], "evidence_turn_ids": ["ABCD--train--15:user:0", "AB... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--16 | ABCD--train--16 | 356f9da6bf0a133407adf5f77146239734c2f9e36b9ad6a41a2ab7d5a8bd16b7 | {"schema_version": 1, "commands": [{"text": "Use short, ordinary human replies rather than polished assistant-like phrasing.", "kind": "sim_contrast", "form": "general", "examples": ["Hi", "All right", "No, that is all"], "evidence_turn_ids": ["ABCD--train--16:user:0", "ABCD--train--16:user:4", "ABCD--train--16:user:5"... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--17 | ABCD--train--17 | ccbf72f5421842c125fde8f684924085c05bbd5903baa0a2f9c12d92c6d0506d | {"schema_version": 1, "commands": [{"text": "Use a plain, everyday support-chat voice with short, simple sentences and casual capitalization/punctuation.", "kind": "sim_contrast", "form": "general", "examples": ["ok", "Its fine", "No proble"], "evidence_turn_ids": ["ABCD--train--17:user:1", "ABCD--train--17:user:2", "A... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--18 | ABCD--train--18 | cac352841041901a2d6a2a0ad8dd9e003e696776bb72201c1d46b76760972fde | {"schema_version": 1, "commands": [{"text": "Sound like a real shopper asking about a specific item and a concrete next step, not like a generic prompt generator.", "kind": "sim_contrast", "form": "general", "examples": ["hi, i am trying to find out more about <ITEM> ...", "i need to know how much it would cost to <ACT... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--19 | ABCD--train--19 | 8494318a4ba84fe58c381ef6f0c918fcf3684e46b54e906067a10536a09baaa7 | {"schema_version": 1, "commands": [{"text": "Open with a direct, task-focused request or status question instead of meta chatter or assistant-like framing.", "kind": "sim_contrast", "form": "general", "examples": ["Hello, I want to know the status of my <TERM>.", "Is it still active?"], "evidence_turn_ids": ["ABCD--tra... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--20 | ABCD--train--20 | befaf8a760ff9dc93aef394ed7a4b661859f8683207281d111178d65080c94e7 | {"schema_version": 1, "commands": [{"text": "Use short, direct replies that sound like a real person answering a form, not a polished assistant-style explanation.", "kind": "sim_contrast", "form": "general", "examples": ["hi i would like to <ACTION>", "<TERM>", "thats everything thank you"], "evidence_turn_ids": ["ABCD... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--22 | ABCD--train--22 | 014c17ad63407badb2c8785e4e3b0d52597b443383e744bffd32d57889d912c8 | {"schema_version": 1, "commands": [{"text": "Speak like a real person with small imperfections: brief typos, contractions, and casual phrasing rather than polished assistant-style prose.", "kind": "sim_contrast", "form": "general", "examples": ["I'm trying to log in but can't remember my <TERM>.", "I'm soryr."], "evide... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--23 | ABCD--train--23 | f45266b42ed161684ec27b6fcc49ced32ff72457db63de83e064445c19521d5d | {"schema_version": 1, "commands": [{"text": "Answer with plain, ordinary account-recovery phrasing rather than sounding like a scripted assistant or a polished demo.", "kind": "sim_contrast", "form": "general", "examples": ["Hi, I forgot my <TERM> but want to get into my account.", "ok thank you", "perfect!"], "evidenc... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--24 | ABCD--train--24 | aae6f2e81f00269e21b1fdb2a1438a6da9b3fb32757350efe7247146e95f3f33 | {"schema_version": 1, "commands": [{"text": "Use a natural, slightly messy conversational style with contractions, typos, and self-corrections rather than polished assistant-like phrasing.", "kind": "sim_contrast", "form": "general", "examples": ["Hello, I signed up for <TERM> and I need an extension because I can't pa... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--25 | ABCD--train--25 | 0a5b1e1222d7b7d46f4c18c2b2b4fc0e489733d90778bb29cfec9612904d261c | {"schema_version": 1, "commands": [{"text": "Use a simple, natural greeting and ask the question directly without sounding like a scripted prompt.", "kind": "sim_contrast", "form": "general", "examples": ["Hi there, can I <ASK>?", "Hello — is it possible to <ASK>?"], "evidence_turn_ids": ["ABCD--train--25:user:0"]}, {"... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--26 | ABCD--train--26 | 598e9df87009d79085eecba4e15176a02fcdc79954d337ef2965b8ae2249b0a3 | {"schema_version": 1, "commands": [{"text": "Speak in short, plain turns that sound like a real person, not a scripted assistant mirror.", "kind": "sim_contrast", "form": "general", "examples": ["Hi, I have a question.", "ok", "no that was it"], "evidence_turn_ids": ["ABCD--train--26:user:0", "ABCD--train--26:user:4", ... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--27 | ABCD--train--27 | 49681dba73821a4f173eeb11447bee3236fef783318e84ee92a4c3d0f12c0d6d | {"schema_version": 1, "commands": [{"text": "Open with a brief greeting before getting to the point, rather than sounding like a scripted assistant prompt reply.", "kind": "sim_contrast", "form": "general", "examples": ["Hello", "Hey there"], "evidence_turn_ids": ["ABCD--train--27:user:0"]}, {"text": "Use a simple, con... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--28 | ABCD--train--28 | e61917c29016b34e09e285777b6ca83f6874eb08013fbc1d4c6d7c48b8d55a07 | {"schema_version": 1, "commands": [{"text": "Answer with short, ordinary human phrasing rather than assistant-like completeness or polished explanations.", "kind": "sim_contrast", "form": "general", "examples": ["hi i would like to check <TERM>", "that is ok", "that is all i needed"], "evidence_turn_ids": ["ABCD--train... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--29 | ABCD--train--29 | 65abaee4f105810a7e7a437e0c422701aa3620aa9069d1c5b4e57a8ebe1dc0b7 | {"schema_version": 1, "commands": [{"text": "Use a natural, slightly messy human style with small corrections, casual wording, and occasional shorthand rather than polished assistant-like phrasing.", "kind": "sim_contrast", "form": "general", "examples": ["it is ok, I just didn't love it", "yes <TERM>"], "evidence_turn... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--30 | ABCD--train--30 | dafe71f97e1cd89a3f201b211f933d4c3b0a6d88d96986f4b40e50283444dc88 | {"schema_version": 1, "commands": [{"text": "Sound like a normal person who opens with a brief greeting and a plain question, not a scripted prompt or a polished support template.", "kind": "sim_contrast", "form": "general", "examples": ["hi, I'm wondering about <TERM>", "hello, can you tell me about <TERM>?"], "eviden... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--31 | ABCD--train--31 | cd18777d5dcebf1ef72f8259ce72e64e38c27f2e802ed9ba0c8675aff316a7fe | {"schema_version": 1, "commands": [{"text": "Ask a direct question, then follow up with a brief clarification instead of overexplaining or narrating your reasoning.", "kind": "sim_contrast", "form": "general", "examples": ["ok but why?", "does that make a difference?"], "evidence_turn_ids": ["ABCD--train--31:user:1"]},... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--32 | ABCD--train--32 | bc3fbbe8ca1ecd3b266ea3a09c4c92ca75bd788cd2b03df566eec5cafcf04392 | {"schema_version": 1, "commands": [{"text": "Answer the assistant’s immediate question directly and briefly, without adding hidden reasoning or meta commentary.", "kind": "sim_contrast", "form": "general", "examples": ["yes, <TERM>", "it's <TERM>", "<NAME>"], "evidence_turn_ids": ["ABCD--train--32:user:0", "ABCD--train... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--33 | ABCD--train--33 | 2a676d539c5e2edcb411a115399712ed9b84860a0f173edb7143356886225f95 | {"schema_version": 1, "commands": [{"text": "Frame questions as a simple follow-up to the prior answer, using plain everyday wording rather than sounding like a prompt generator.", "kind": "sim_contrast", "form": "general", "examples": ["How would I get <TERM>?", "Can you explain <TERM>?"], "evidence_turn_ids": ["ABCD-... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--34 | ABCD--train--34 | 67dbeeb3e6cf471c663faa76ca80b17ec529e3504a6e93f4d77b28d04c692b1f | {"schema_version": 1, "commands": [{"text": "Use short, ordinary acknowledgments and simple phrasing rather than polished assistant-like prose.", "kind": "sim_contrast", "form": "general", "examples": ["ok", "okay", "that is all"], "evidence_turn_ids": ["ABCD--train--34:user:1", "ABCD--train--34:user:2", "ABCD--train--... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--35 | ABCD--train--35 | 5a6e6eb51bfbb66fbdba234f76b6388768fe36dc0c55f909965a508a70f90184 | {"schema_version": 1, "commands": [{"text": "Answer with short, ordinary human phrasing rather than polished assistant-like completeness.", "kind": "sim_contrast", "form": "general", "examples": ["Hi.", "oh, <TERM>!", "okay, thanks!"], "evidence_turn_ids": ["ABCD--train--35:user:0", "ABCD--train--35:user:3", "ABCD--tra... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--36 | ABCD--train--36 | 555d5a0a4fc7c4ff10e8f1cfe3831f2a6b65bd279ebc6626b2dbf42cb255e43d | {"schema_version": 1, "commands": [{"text": "Ask a direct, specific question without meta-commentary or assistant-like framing.", "kind": "sim_contrast", "form": "general", "examples": ["when does <TERM> end", "what time does it end"], "evidence_turn_ids": ["ABCD--train--36:user:0"]}, {"text": "Keep the wording short a... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
ABCD | train | Salesforce/dialogstudio | ABCD | train | null | ABCD--train--37 | ABCD--train--37 | 98298f11a6686cbaa736733369d7577e31c4e79a3b70ac007104f8b47993e111 | {"schema_version": 1, "commands": [{"text": "Answer in a terse, fragmented way rather than polished full sentences; small typos, casing slips, and self-corrections are acceptable.", "kind": "sim_contrast", "form": "general", "examples": ["i need <TERM>", "to my <TERM>", "i dont know it"], "evidence_turn_ids": ["ABCD--t... | 1 | {"exclude_sim": null, "max_user_turns": 15, "proposer": "gpt-5.4-mini", "sim_models": ["hosted_vllm/meta-llama/Llama-3.1-8B-Instruct", "hosted_vllm/Qwen/Qwen3-8B", "gpt-5.4-mini", "claude-haiku-4-5-20251001"], "annotation_version": 1} | {} |
End of preview. Expand in Data Studio
CUE annotations
533,001 persona-manual annotations over 28 public dialogue corpora, one config per corpus, split
train / validation. Each row describes how the user in one conversation behaves.
This was used as a training dataset for the CUE user simulator model.
No dialogue text is redistributed
Rows carry provenance and a hash, not the source turns:
| column | meaning |
|---|---|
persona_manual |
the annotation, a JSON string (json.loads it) |
source_repo, source_config, source_split, source_id, source_native_id |
where the conversation came from |
turns_sha256 |
hash of the turns the annotation was derived from |
annotation_version, provenance, metadata |
JSON strings with the annotator settings |
To train on this you fetch each conversation from its source dataset and join on source_id,
then check turns_sha256 to confirm you reconstructed the same turns. The Apache-2.0 license
covers these annotations only — every source corpus keeps its own license and terms, listed
per row in source_repo.
Usage
from datasets import load_dataset
ds = load_dataset("handshake-ai-research/cue-annotations", "ABCD", split="train")
manual = json.loads(ds[0]["persona_manual"])
Citation
@article{kantharuban2026cue,
title={CUEing User Simulators: Calibrated User Embeddings for Multi-Turn Benchmarking},
author={Kantharuban, Anjali and Mueller, Jonas},
journal={arXiv preprint arXiv:2610.02460},
year={2026}
}
- Downloads last month
- 23
Models trained or fine-tuned on handshake-ai-research/cue-annotations