WorkflowEvals
Collection
https://github.com/typesafe-ai/WorkflowEvals • https://evals.typesafe.ai • 4 items • Updated • 3
case_id large_stringlengths 20 45 | input_json large_stringlengths 569 1.96k | metadata_json large_stringlengths 90 363 | openai dict | anthropic dict | consensus dict |
|---|---|---|---|---|---|
access-2fa-old-number-flight/t0 | {"case_id":"access-2fa-old-number-flight/t0","dialog_id":"access-2fa-old-number-flight","turn":0,"transcript":[{"speaker":"customer","text":"I changed phone numbers last month and now the login code still goes to my old number. I fly out tomorrow morning and need the app to keep an eye on the card while I'm away. How d... | {"dialog_id":"access-2fa-old-number-flight","turn":0,"theme":"access","note":"2FA phone changed, travel tomorrow: urgency 'pressing'. B raises priority; C high tier sets urgent; all give account_recovery_steps."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
"arguments_json": "{\"intent\":\"account... |
access-2fa-old-number-flight/t1 | {"case_id":"access-2fa-old-number-flight/t1","dialog_id":"access-2fa-old-number-flight","turn":1,"transcript":[{"speaker":"customer","text":"I changed phone numbers last month and now the login code still goes to my old number. I fly out tomorrow morning and need the app to keep an eye on the card while I'm away. How d... | {"dialog_id":"access-2fa-old-number-flight","turn":1,"theme":"access","note":"Quoted regulator suggestion plus explicit negation: threat_legal_regulatory should be no. Calm. Expect recovery steps, not compliance."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
"arguments_json": "{\"intent\":\"account... |
access-2fa-old-number-flight/t2 | {"case_id":"access-2fa-old-number-flight/t2","dialog_id":"access-2fa-old-number-flight","turn":2,"transcript":[{"speaker":"customer","text":"I changed phone numbers last month and now the login code still goes to my old number. I fly out tomorrow morning and need the app to keep an eye on the card while I'm away. How d... | {"dialog_id":"access-2fa-old-number-flight","turn":2,"theme":"access","note":"Still waiting on recovery, adds a general question: issue_resolved should be no. Intent stays account_access vs general_question; either way an explanation, not a close."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
"arguments_json": "{\"intent\":\"account... |
access-app-spinner-indirect-human/t0 | {"case_id":"access-app-spinner-indirect-human/t0","dialog_id":"access-app-spinner-indirect-human","turn":0,"transcript":[{"speaker":"customer","text":"The app has been stuck on the loading spinner for two days. Deleted it, reinstalled, same thing. The website says 'session expired' every time I try. iPhone if that matt... | {"dialog_id":"access-app-spinner-indirect-human","turn":0,"theme":"access","note":"App not loading, concrete details, no time pressure. Intent account_access vs general_question; expect recovery steps or an explanation, no escalation."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
"arguments_json": "{\"intent\":\"account... |
access-app-spinner-indirect-human/t1 | {"case_id":"access-app-spinner-indirect-human/t1","dialog_id":"access-app-spinner-indirect-human","turn":1,"transcript":[{"speaker":"customer","text":"The app has been stuck on the loading spinner for two days. Deleted it, reinstalled, same thing. The website says 'session expired' every time I try. iPhone if that matt... | {"dialog_id":"access-app-spinner-indirect-human","turn":1,"theme":"access","note":"Indirect request for a person, frustration around level 2. A (0.5) and C high tier hand off; B needs requests_human > 0.7 and may keep the conversation."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"hand... |
access-code-error-self-resolved/t0 | {"case_id":"access-code-error-self-resolved/t0","dialog_id":"access-code-error-self-resolved","turn":0,"transcript":[{"speaker":"customer","text":"hey I cant get into the app, it keeps saying 'something went wrong' after I put in the code. tried like 6 times over the last hour. no big deal but annoying, I wanna pay my ... | {"dialog_id":"access-code-error-self-resolved","turn":0,"theme":"access","note":"Repeated failed attempts and a due date four days out: urgency between 1 and 2, B's E[urgency] > 1.5 is the close call. Intent account_access; expect recovery steps."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
"arguments_json": "{\"intent\":\"account... |
access-code-error-self-resolved/t1 | {"case_id":"access-code-error-self-resolved/t1","dialog_id":"access-code-error-self-resolved","turn":1,"transcript":[{"speaker":"customer","text":"hey I cant get into the app, it keeps saying 'something went wrong' after I put in the code. tried like 6 times over the last hour. no big deal but annoying, I wanna pay my ... | {"dialog_id":"access-code-error-self-resolved","turn":1,"theme":"access","note":"Resolved itself mid-thread, calm: auto-close everywhere, including C's high tier which needs almost no frustration mass."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"clos... |
access-declined-gas-station-child/t0 | {"case_id":"access-declined-gas-station-child/t0","dialog_id":"access-declined-gas-station-child","turn":0,"transcript":[{"speaker":"customer","text":"CARD DECLINED AT THE GAS STATION. im at the pump with my 4 year old in the back seat, quarter tank, 40 miles from home, its 9pm. I have NOT lost the card, its in my hand... | {"dialog_id":"access-declined-gas-station-child","turn":0,"theme":"access","note":"Stranded now with a child: urgency 3. A sets urgent and hands off; B raises priority and explains; C low tier sets urgent and explains. Caps read as stress, frustration around 2."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"hand... |
access-declined-gas-station-child/t1 | {"case_id":"access-declined-gas-station-child/t1","dialog_id":"access-declined-gas-station-child","turn":1,"transcript":[{"speaker":"customer","text":"CARD DECLINED AT THE GAS STATION. im at the pump with my 4 year old in the back seat, quarter tank, 40 miles from home, its 9pm. I have NOT lost the card, its in my hand... | {"dialog_id":"access-declined-gas-station-child","turn":1,"theme":"access","note":"Went through, so partly resolved, but anger at the system with undirected profanity: frustration 3, not 4. A hands off (P(>=3) > 0.35); B empathizes then may close; C low tier's 0.7 handoff bar is the close call."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"hand... |
access-declined-lisbon-hotel/t0 | {"case_id":"access-declined-lisbon-hotel/t0","dialog_id":"access-declined-lisbon-hotel","turn":0,"transcript":[{"speaker":"customer","text":"Good evening. I am in Lisbon for work and my card was refused at two restaurants and one ATM today. Before leaving I did not inform you of travel, perhaps this is the reason? I wo... | {"dialog_id":"access-declined-lisbon-hotel","turn":0,"theme":"access","note":"Declined abroad, formal non-native register, payment due Friday: urgency 2. B sets high priority; C low tier does not (needs P(>=3) > 0.6). All explain."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"expl... |
access-declined-lisbon-hotel/t1 | {"case_id":"access-declined-lisbon-hotel/t1","dialog_id":"access-declined-lisbon-hotel","turn":1,"transcript":[{"speaker":"customer","text":"Good evening. I am in Lisbon for work and my card was refused at two restaurants and one ATM today. Before leaving I did not inform you of travel, perhaps this is the reason? I wo... | {"dialog_id":"access-declined-lisbon-hotel","turn":1,"theme":"access","note":"Fixed, with a competitor comparison and an explicit negation of switching: churn level 1, not 2, so B's retention offer (P(>=2) > 0.5) should not fire. Resolved but not fully calm."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"clos... |
access-declined-lisbon-hotel/t2 | {"case_id":"access-declined-lisbon-hotel/t2","dialog_id":"access-declined-lisbon-hotel","turn":2,"transcript":[{"speaker":"customer","text":"Good evening. I am in Lisbon for work and my card was refused at two restaurants and one ATM today. Before leaving I did not inform you of travel, perhaps this is the reason? I wo... | {"dialog_id":"access-declined-lisbon-hotel","turn":2,"theme":"access","note":"New general question after the decline is resolved: issue_resolved no, intent drifts to general_question. Expect explain_charge_or_policy, no close."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"expl... |
access-declined-online-buried-vape/t0 | {"case_id":"access-declined-online-buried-vape/t0","dialog_id":"access-declined-online-buried-vape","turn":0,"transcript":[{"speaker":"customer","text":"Trying to pay my kid's school lunch account online and the card keeps getting declined, three tries. It was fine last week. Also there's some $62.50 thing from a vape ... | {"dialog_id":"access-declined-online-buried-vape","turn":0,"theme":"access","note":"Decline story with an unrecognised charge buried in it: intent card_declined by stated outcome, but reports_unauthorized should clear 0.5. A and C ask about possession; B disputes at once."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
"arguments_json": "{\"intent\":\"card_de... |
access-declined-online-buried-vape/t1 | {"case_id":"access-declined-online-buried-vape/t1","dialog_id":"access-declined-online-buried-vape","turn":1,"transcript":[{"speaker":"customer","text":"Trying to pay my kid's school lunch account online and the card keeps getting declined, three tries. It was fine last week. Also there's some $62.50 thing from a vape ... | {"dialog_id":"access-declined-online-buried-vape","turn":1,"theme":"access","note":"Card in hand, charge on file, unauthorized now confirmed. A proposes a freeze; B opens a dispute; C proposes a dispute."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
"arguments_json": "{\"intent\":\"card_de... |
access-declined-online-buried-vape/t2 | {"case_id":"access-declined-online-buried-vape/t2","dialog_id":"access-declined-online-buried-vape","turn":2,"transcript":[{"speaker":"customer","text":"Trying to pay my kid's school lunch account online and the card keeps getting declined, three tries. It was fine last week. Also there's some $62.50 thing from a vape ... | {"dialog_id":"access-declined-online-buried-vape","turn":2,"theme":"access","note":"Resigned 'sure, whatever' to the pending freeze: accepts, but confidence is the test. B (0.5) executes, A (0.6) likely, C high tier needs 0.75."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "freeze_card",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "freeze_card",
... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "freeze_card",
"arguments_json": "{\"card_id\":\"card-... |
access-declined-self-frozen/t0 | {"case_id":"access-declined-self-frozen/t0","dialog_id":"access-declined-self-frozen","turn":0,"transcript":[{"speaker":"customer","text":"my card got declined at the pharmacy just now. tried twice. its embarrassing. whats going on"}],"customer":{"tenure_months":22,"vip":false,"prior_contacts_30d":0,"prior_refunds_90d"... | {"dialog_id":"access-declined-self-frozen","turn":0,"theme":"access","note":"Decline complaint, card on file is frozen. Intent card_declined, frustration around 1-2, urgency 'soon'. Expect explain_charge_or_policy."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"expl... |
access-declined-self-frozen/t1 | {"case_id":"access-declined-self-frozen/t1","dialog_id":"access-declined-self-frozen","turn":1,"transcript":[{"speaker":"customer","text":"my card got declined at the pharmacy just now. tried twice. its embarrassing. whats going on"},{"speaker":"assistant","text":"Sorry about that. I've had a look and the card ending 8... | {"dialog_id":"access-declined-self-frozen","turn":1,"theme":"access","note":"Card found, in hand, nothing unauthorized: reports_unauthorized no. Medication today sits between urgency 2 and 3; A's stranded rule (P(>=3) > 0.4) is the close call."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"expl... |
access-declined-self-frozen/t2 | {"case_id":"access-declined-self-frozen/t2","dialog_id":"access-declined-self-frozen","turn":2,"transcript":[{"speaker":"customer","text":"my card got declined at the pharmacy just now. tried twice. its embarrassing. whats going on"},{"speaker":"assistant","text":"Sorry about that. I've had a look and the card ending 8... | {"dialog_id":"access-declined-self-frozen","turn":2,"theme":"access","note":"Resolved and calm: auto-close everywhere."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"clos... |
access-expired-card-never-arrived/t0 | {"case_id":"access-expired-card-never-arrived/t0","dialog_id":"access-expired-card-never-arrived","turn":0,"transcript":[{"speaker":"customer","text":"My card was declined at the grocery store, the cashier said it's expired. It does say 08/26 on it. I never got a new one in the mail. Money's a bit tight this month so I... | {"dialog_id":"access-expired-card-never-arrived","turn":0,"theme":"access","note":"Card expired because the replacement never arrived: card_declined vs delivery_issue, with an order on file. Hardship level 1, urgency 2. Either intent leads to an explanation or the order status."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"hand... |
access-expired-card-never-arrived/t1 | {"case_id":"access-expired-card-never-arrived/t1","dialog_id":"access-expired-card-never-arrived","turn":1,"transcript":[{"speaker":"customer","text":"My card was declined at the grocery store, the cashier said it's expired. It does say 08/26 on it. I never got a new one in the mail. Money's a bit tight this month so I... | {"dialog_id":"access-expired-card-never-arrived","turn":1,"theme":"access","note":"Yes to a reissue that is not a PENDING action, so nothing executes on consent. Hypothetical churn ('might have to think about') sits between levels 1 and 2 for B's retention rule; hardship rises to 2-3 (late on rent)."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
"arguments_json": "{\"intent\":\"deliver... |
access-expired-card-never-arrived/t2 | {"case_id":"access-expired-card-never-arrived/t2","dialog_id":"access-expired-card-never-arrived","turn":2,"transcript":[{"speaker":"customer","text":"My card was declined at the grocery store, the cashier said it's expired. It does say 08/26 on it. I never got a new one in the mail. Money's a bit tight this month so I... | {"dialog_id":"access-expired-card-never-arrived","turn":2,"theme":"access","note":"Calm follow-up question while waiting on the card: not resolved, general_question drift. Expect an explanation, no close."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
"arguments_json": "{\"intent\":\"deliver... |
access-kid-passcode-last-four/t0 | {"case_id":"access-kid-passcode-last-four/t0","dialog_id":"access-kid-passcode-last-four","turn":0,"transcript":[{"speaker":"customer","text":"Locked out of the app after my kid mashed the wrong passcode like 5 times. My card is the one ending in 4471 if you need it. Would be nice to have it back before the weekend, no... | {"dialog_id":"access-kid-passcode-last-four","turn":0,"theme":"access","note":"Only a last-four is shared: shares_credentials must be no. Urgency 'soon'. Expect account_recovery_steps."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "set_intent",
"arguments_json": "{\"intent\":\"account... |
access-kid-passcode-last-four/t1 | {"case_id":"access-kid-passcode-last-four/t1","dialog_id":"access-kid-passcode-last-four","turn":1,"transcript":[{"speaker":"customer","text":"Locked out of the app after my kid mashed the wrong passcode like 5 times. My card is the one ending in 4471 if you need it. Would be nice to have it back before the weekend, no... | {"dialog_id":"access-kid-passcode-last-four","turn":1,"theme":"access","note":"Declines an offer that is not a PENDING proposal and signals done; calm. Close in every policy."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"clos... |
access-locked-out-full-password/t0 | {"case_id":"access-locked-out-full-password/t0","dialog_id":"access-locked-out-full-password","turn":0,"transcript":[{"speaker":"customer","text":"I've been locked out of the app since this morning, it says too many attempts. My password is Sunflower2019! and I'm certain that's right, so I don't know why it won't take ... | {"dialog_id":"access-locked-out-full-password","turn":0,"theme":"access","note":"Full password typed into chat: every policy redacts and warns first, before any recovery help. Urgency low, intent account_access."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "redact_transcript",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "redact_transcr... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "redact_transcript",
"arguments_json": "{}"
},... |
access-locked-out-full-password/t1 | {"case_id":"access-locked-out-full-password/t1","dialog_id":"access-locked-out-full-password","turn":1,"transcript":[{"speaker":"customer","text":"I've been locked out of the app since this morning, it says too many attempts. My password is Sunflower2019! and I'm certain that's right, so I don't know why it won't take ... | {"dialog_id":"access-locked-out-full-password","turn":1,"theme":"access","note":"Consent to a reset link, which is not a PENDING action, so nothing executes; urgency 'would like it soon'. Expect account_recovery_steps."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "redact_transcript",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "redact_transcr... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "redact_transcript",
"arguments_json": "{}"
},... |
access-locked-out-full-password/t2 | {"case_id":"access-locked-out-full-password/t2","dialog_id":"access-locked-out-full-password","turn":2,"transcript":[{"speaker":"customer","text":"I've been locked out of the app since this morning, it says too many attempts. My password is Sunflower2019! and I'm certain that's right, so I don't know why it won't take ... | {"dialog_id":"access-locked-out-full-password","turn":2,"theme":"access","note":"Resolved and calm: auto-close in every policy."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "redact_transcript",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "redact_transcr... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "redact_transcript",
"arguments_json": "{}"
},... |
access-reset-link-never-worked/t0 | {"case_id":"access-reset-link-never-worked/t0","dialog_id":"access-reset-link-never-worked","turn":0,"transcript":[{"speaker":"customer","text":"Second time here. Yesterday your bot sent me a password reset link and it just went to a page saying 'token invalid'. Still can't get in. Not angry, just want it to actually w... | {"dialog_id":"access-reset-link-never-worked","turn":0,"theme":"access","note":"Alleges the assistant's earlier reset link failed: claims_agent_error yes. A hands off to a supervisor; B and C apologize and continue. Frustration near level 2 despite 'not angry'; C high tier P(>=2)>0.4 is the close call."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human"... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
"arguments_json": "{\"queue\":\"super... |
access-reset-link-never-worked/t1 | {"case_id":"access-reset-link-never-worked/t1","dialog_id":"access-reset-link-never-worked","turn":1,"transcript":[{"speaker":"customer","text":"Second time here. Yesterday your bot sent me a password reset link and it just went to a page saying 'token invalid'. Still can't get in. Not angry, just want it to actually w... | {"dialog_id":"access-reset-link-never-worked","turn":1,"theme":"access","note":"'That fixed it, BUT': resolved with a mild complaint. A and B close; C should mark resolved but leave the ticket open since not all the mass is calm."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human"... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
"arguments_json": "{\"queue\":\"super... |
access-reset-link-never-worked/t2 | {"case_id":"access-reset-link-never-worked/t2","dialog_id":"access-reset-link-never-worked","turn":2,"transcript":[{"speaker":"customer","text":"Second time here. Yesterday your bot sent me a password reset link and it just went to a page saying 'token invalid'. Still can't get in. Not angry, just want it to actually w... | {"dialog_id":"access-reset-link-never-worked","turn":2,"theme":"access","note":"Stale double-charge (resolved, past) vs a current fee question: intent shifts toward billing_dispute or general_question, claims_agent_error should stay low. Fee on file, no hardship, so no waiver."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human"... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
"arguments_json": "{\"queue\":\"super... |
billing-double-charge-152/t0 | {"case_id":"billing-double-charge-152/t0","dialog_id":"billing-double-charge-152","turn":0,"transcript":[{"speaker":"customer","text":"I WAS CHARGED TWICE FOR THE SAME $152 ORDER FROM NORTHWIND ON THE 3RD. TWO IDENTICAL CHARGES. THIS IS THE SECOND TIME THIS YEAR. FIX IT."}],"customer":{"tenure_months":31,"vip":false,"p... | {"dialog_id":"billing-double-charge-152","turn":0,"theme":"billing","note":"All-caps anger at the situation, no insult: frustration 3, not 4. A hands off on anger; C mid hands off as too frustrated; B empathises, raises priority and, at $152, sends the refund to a person."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"hand... |
billing-double-charge-152/t1 | {"case_id":"billing-double-charge-152/t1","dialog_id":"billing-double-charge-152","turn":1,"transcript":[{"speaker":"customer","text":"I WAS CHARGED TWICE FOR THE SAME $152 ORDER FROM NORTHWIND ON THE 3RD. TWO IDENTICAL CHARGES. THIS IS THE SECOND TIME THIS YEAR. FIX IT."},{"speaker":"assistant","text":"I'm really sorr... | {"dialog_id":"billing-double-charge-152","turn":1,"theme":"billing","note":"Sarcastic acceptance. If read as accepts, the amount straddles the limits: A (over $50) and B (over $150) hand off to refunds, C mid (under $200) pays it."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "clear_pending",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "clear_pending"... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "clear_pending",
"arguments_json": "{}"
},
... |
billing-double-charge-152/t2 | {"case_id":"billing-double-charge-152/t2","dialog_id":"billing-double-charge-152","turn":2,"transcript":[{"speaker":"customer","text":"I WAS CHARGED TWICE FOR THE SAME $152 ORDER FROM NORTHWIND ON THE 3RD. TWO IDENTICAL CHARGES. THIS IS THE SECOND TIME THIS YEAR. FIX IT."},{"speaker":"assistant","text":"I'm really sorr... | {"dialog_id":"billing-double-charge-152","turn":2,"theme":"billing","note":"'That fixed it, BUT': not resolved. The February complaint is stale history, not a new claim. 'Anyone who can actually look into it' is an indirect request for a person, near the requests_human threshold."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human"... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
"arguments_json": "{\"queue\":\"super... |
billing-double-charge-coffee/t0 | {"case_id":"billing-double-charge-coffee/t0","dialog_id":"billing-double-charge-coffee","turn":0,"transcript":[{"speaker":"customer","text":"hey, order 71133 from ridgeline coffee got charged twice, two lines of $24.50 on sept 2, i only ordered once. can i get one of them back"}],"customer":{"tenure_months":14,"vip":fa... | {"dialog_id":"billing-double-charge-coffee","turn":0,"theme":"billing","note":"Duplicate charge, $24.50 eligible: under B's $25 instant limit and C mid's $30 automatic limit; A proposes. refund_reason duplicate_or_erroneous_charge should be confident."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
"arguments_json": "{\"queue\":\"super... |
billing-double-charge-coffee/t1 | {"case_id":"billing-double-charge-coffee/t1","dialog_id":"billing-double-charge-coffee","turn":1,"transcript":[{"speaker":"customer","text":"hey, order 71133 from ridgeline coffee got charged twice, two lines of $24.50 on sept 2, i only ordered once. can i get one of them back"},{"speaker":"assistant","text":"You're ri... | {"dialog_id":"billing-double-charge-coffee","turn":1,"theme":"billing","note":"Yes to a different amount plus a new request: proposal_reply should be unclear_or_conditional, not accepts; the $24.50 must not fire. Intent drifts toward refund_request / changed_mind."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
"arguments_json": "{\"queue\":\"super... |
billing-double-charge-coffee/t2 | {"case_id":"billing-double-charge-coffee/t2","dialog_id":"billing-double-charge-coffee","turn":2,"transcript":[{"speaker":"customer","text":"hey, order 71133 from ridgeline coffee got charged twice, two lines of $24.50 on sept 2, i only ordered once. can i get one of them back"},{"speaker":"assistant","text":"You're ri... | {"dialog_id":"billing-double-charge-coffee","turn":2,"theme":"billing","note":"Clear acceptance of the pending $24.50 refund after the detour: A, B and C all execute it and mark resolved."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "clear_pending",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "clear_pending"... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "clear_pending",
"arguments_json": "{}"
},
... |
billing-fee-already-waived/t0 | {"case_id":"billing-fee-already-waived/t0","dialog_id":"billing-fee-already-waived","turn":0,"transcript":[{"speaker":"customer","text":"Hi, could you waive the $29 late fee from August? I paid two days late, first time in almost three years."}],"customer":{"tenure_months":33,"vip":false,"prior_contacts_30d":0,"prior_r... | {"dialog_id":"billing-fee-already-waived","turn":0,"theme":"billing","note":"Politely asks for a waiver the record shows was already credited: the already-refunded rule in every policy explains rather than pays twice."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"expl... |
billing-fee-already-waived/t1 | {"case_id":"billing-fee-already-waived/t1","dialog_id":"billing-fee-already-waived","turn":1,"transcript":[{"speaker":"customer","text":"Hi, could you waive the $29 late fee from August? I paid two days late, first time in almost three years."},{"speaker":"assistant","text":"Good news: that $29 late fee was already rev... | {"dialog_id":"billing-fee-already-waived","turn":1,"theme":"billing","note":"Resolved and calm: auto-close in every policy."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"clos... |
billing-foreign-fee-courtesy/t0 | {"case_id":"billing-foreign-fee-courtesy/t0","dialog_id":"billing-foreign-fee-courtesy","turn":0,"transcript":[{"speaker":"customer","text":"Good afternoon. I booked a hotel in Lisbon for $420 on 27 August and was charged a $12.60 foreign transaction fee. The booking was priced and paid in US dollars, so I fail to see ... | {"dialog_id":"billing-foreign-fee-courtesy","turn":0,"theme":"billing","note":"Formal; recognises the charge and disputes only the fee: billing_dispute, reports_unauthorized no. Claim on file is not eligible; B's goodwill refund needs refund_reason confidence > 0.8, which this reason barely fits."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human"... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
"arguments_json": "{\"queue\":\"super... |
billing-foreign-fee-courtesy/t1 | {"case_id":"billing-foreign-fee-courtesy/t1","dialog_id":"billing-foreign-fee-courtesy","turn":1,"transcript":[{"speaker":"customer","text":"Good afternoon. I booked a hotel in Lisbon for $420 on 27 August and was charged a $12.60 foreign transaction fee. The booking was priced and paid in US dollars, so I fail to see ... | {"dialog_id":"billing-foreign-fee-courtesy","turn":1,"theme":"billing","note":"Negated regulator mention: threat_legal_regulatory should be no. Mild irritation, money_back with no hardship. Expect explain-not-eligible, no flag, no handoff."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"expl... |
billing-foreign-fee-courtesy/t2 | {"case_id":"billing-foreign-fee-courtesy/t2","dialog_id":"billing-foreign-fee-courtesy","turn":2,"transcript":[{"speaker":"customer","text":"Good afternoon. I booked a hotel in Lisbon for $420 on 27 August and was charged a $12.60 foreign transaction fee. The booking was priced and paid in US dollars, so I fail to see ... | {"dialog_id":"billing-foreign-fee-courtesy","turn":2,"theme":"billing","note":"Resolved but not calm: A and B close the ticket; C should mark resolved and leave it open because frustration mass sits above calm."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "respond_with",
"arguments_json": "{\"message\":\"clos... |
billing-interest-after-full-payment/t0 | {"case_id":"billing-interest-after-full-payment/t0","dialog_id":"billing-interest-after-full-payment","turn":0,"transcript":[{"speaker":"customer","text":"Quick one. Last year you charged me a late fee that shouldn't have existed; it got fixed after a few weeks, so that's done. Today's question is separate: why is ther... | {"dialog_id":"billing-interest-after-full-payment","turn":0,"theme":"billing","note":"Stale, already-fixed complaint plus a current question. Intent should be general_question (explanation_only), not billing_dispute; nothing on file to refund; claims_agent_error refers to the past, so no."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human"... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
"arguments_json": "{\"queue\":\"super... |
billing-interest-after-full-payment/t1 | {"case_id":"billing-interest-after-full-payment/t1","dialog_id":"billing-interest-after-full-payment","turn":1,"transcript":[{"speaker":"customer","text":"Quick one. Last year you charged me a late fee that shouldn't have existed; it got fixed after a few weeks, so that's done. Today's question is separate: why is ther... | {"dialog_id":"billing-interest-after-full-payment","turn":1,"theme":"billing","note":"Alleges the assistant's answer is wrong (claims_agent_error) and adds concrete details; intent shifts toward billing_dispute. A hands off to a supervisor; B and C flag and apologise while continuing."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human"... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
"arguments_json": "{\"queue\":\"super... |
billing-interest-after-full-payment/t2 | {"case_id":"billing-interest-after-full-payment/t2","dialog_id":"billing-interest-after-full-payment","turn":2,"transcript":[{"speaker":"customer","text":"Quick one. Last year you charged me a late fee that shouldn't have existed; it got fixed after a few weeks, so that's done. Today's question is separate: why is ther... | {"dialog_id":"billing-interest-after-full-payment","turn":2,"theme":"billing","note":"Resolved in thread and the error claim withdrawn. Calm enough to auto-close in every policy; claims_agent_error should now be no."} | {
"status": "ok",
"model": {
"name": "gpt-6-astra",
"provider": "openai",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
... | {
"status": "ok",
"model": {
"name": "claude-fable-5-1",
"provider": "anthropic",
"reasoning_effort": "high",
"question_mode": "one"
},
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human"... | {
"status": "ok",
"method": "mean_probabilities",
"contributors": [
"openai",
"anthropic"
],
"used_opus_fallback": false,
"decisions": [
{
"policy_id": "A",
"status": "ok",
"actions": [
{
"name": "handoff_human",
"arguments_json": "{\"queue\":\"super... |
End of preview. Expand in Data Studio
Customer service
Snapshot: 2026-09-28. 204 cases and 3,287 question instances. Default reference: consensus. Labels are model-generated references.
Data
Load configuration cases, questions, or run_results; all have a test split.
- cases: one row per
case_id, with the complete input ininput_json, descriptivemetadata_json, andopenai,anthropic, andconsensuslabelsets. Decisions are grouped bypolicy_idand containstatus,actions, andprimary_action. - questions: one row per distinct case/node/question/input combination, identified by
question_instance_id.question_jsonandstate_jsondescribe the model input. Each labelset containsanswer_json,probabilities,confidence,confidence_method, andexpected_scorefor score questions.answer_jsonis the modal answer (null for ties); the expected score is a separate numeric value.
Independent labels include the actual model, reasoning effort, question mode, and
used_opus_fallback. Missing labels have explicit status and null values; the case is retained.
Question status not_answered means that reference did not answer that question instance.
Scoring
dataset.json lists policy IDs and the applicable comparison rules:
exact_actions: compare complete action sets including arguments; ignore order and duplicates.primary_action: compare the designated primary action including its arguments.
Customer service has no designated primary action, so primary_action is null. Its consensus blends derived workflow signals (including refined intent and combined claims), so blending raw question answers is not always equivalent to its case reference.
Run results
run_results contains 9 code-route runs selected by the final plots:
1,836 rows, one per run_id and case_id, with 29,204
recorded question answers. The test split contains every exported row; nothing is partitioned.
Each row includes the model's name, provider, reasoning effort, and question mode;
execution status, cost_usd, wall_time_s, summed_call_time_s, token counts, and call count;
and two nested lists:
questions:node_id,question_id,kind, the actualquestion_jsonandstate_json,status,answer_json,probabilities,expected_score,confidence, andconfidence_method. Inputs are repeated so each result is self-contained. Answers are the recorded run values;expected_scoreis calculated from the distribution for score questions. Only recorded questions appear: branches not taken do not produce fabricated answers. Dynamic choice options are preserved exactly, including instances absent from the reference question table.decisions:policy_id,status,actions,primary_action, andscores. Each score hasmetric_idand a booleanvalueagainst the consensus reference (null when inapplicable). Metrics use onlyexact_actionsorprimary_actioncomparisons.
Runs use exactly the consensus scoring cases in cases. Prediction errors remain in scoring
and count as wrong. Measurements are per case, shared across policy decisions; missing
measurements remain null. Cost is recorded in USD using the basis named in each run summary.
dataset.json contains runs, keyed by run_id, with plotted aggregate scores, denominators,
mean cost/time, and measurement counts. The selected run for a model configuration is the one
used by the plot (highest primary accuracy), not necessarily its newest execution. Overall
plot points average workflows equally for configurations present in all four workflows.
Consensus
Question-level consensus averages the available probability distributions for the same question
and input. contributors and weights specify 0.5 each for two sources or 1.0 for a single
source. Consensus confidence is the maximum blended probability.
Case-level consensus contains the final reference actions from applying the workflow to its blended signals. These are the references used for evaluation. Distinct dynamic question instances remain separate in the question table.
Coverage
| Labelset | Cases with decisions | Opus fallback cases |
|---|---|---|
| openai | 204 | 0 |
| anthropic | 204 | 0 |
| consensus | 204 | 0 |
Encoding
Fields ending in _json are JSON-encoded text; decode with json.loads.
Probability distributions are lists of option/probability pairs. Action arguments are
preserved in arguments_json. Descriptive metadata is separate from model-visible state.
- Downloads last month
- 10