Releasing the Inner Jev: A Simple Recipe for Open Decision Models
Paper • 2610.17205 • Published • 2
id stringlengths 15 65 | group_id stringlengths 14 44 | family stringclasses 98
values | part stringclasses 5
values | source_dataset stringclasses 63
values | source_license stringclasses 28
values | target_origin stringclasses 8
values | state stringlengths 2 40.6k | state_is_json bool 2
classes | question stringlengths 15 23.6k | options listlengths 2 16 | qtype stringclasses 3
values | label int64 0 15 | target_distribution listlengths 2 16 ⌀ | tier stringclasses 4
values | language stringclasses 22
values | use_case stringclasses 20
values | source_ref stringlengths 16 185 ⌀ | in_jeff_1_1 bool 2
classes |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
decider/insincere_questions/89::q0 | decider/insincere_questions/30172d6a161268e4 | insincere_questions | decider | SetFit/insincere-questions | unknown | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Assuming the technology for interplanetary migration has been developed, how would the colonization of Mars happen? | false | Is this forum question insincere (rhetorical, provocative, or not seeking a real answer)? | [
{
"id": "insincere",
"description": "insincere"
},
{
"id": "sincere",
"description": "sincere"
}
] | choice | 1 | [
2.8687815431603965e-7,
0.9999997131218457
] | null | null | null | {"task": "insincere_questions", "example": 89, "question": 0} | false |
decider/mnli/91::q0 | decider/mnli/ecd44535a1a96662 | mnli | decider | nyu-mll/multi_nli | mixed (cc-by-3.0, cc-by-sa-3.0, mit, other) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Text 1: It vanished, and the two men were also gone.
Text 2: The men had left long before it vanished. | false | What is the logical relation between text 1 and text 2? | [
{
"id": "entailment (text 2 follows from text 1)",
"description": "entailment (text 2 follows from text 1)"
},
{
"id": "neutral",
"description": "neutral"
},
{
"id": "contradiction",
"description": "contradiction"
}
] | choice | 1 | [
0.0000030139674077391865,
0.9999924812513774,
0.000004504781214868072
] | null | null | null | {"task": "mnli", "example": 91, "question": 0} | true |
decider/tweet_emotion/165::q0 | decider/tweet_emotion/e44a76d174a8ae46 | tweet_emotion | decider | cardiffnlp/tweet_eval | unknown | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | @user #heyday Race war 2016 | false | Which emotion does this tweet express? | [
{
"id": "optimism",
"description": "optimism"
},
{
"id": "anger",
"description": "anger"
},
{
"id": "sadness",
"description": "sadness"
},
{
"id": "joy",
"description": "joy"
}
] | choice | 3 | [
1.4673574388519093e-8,
0.4999999127668892,
6.838677298625467e-9,
0.5000000657208591
] | null | null | null | {"task": "tweet_emotion", "example": 165, "question": 0} | true |
decider/copa/218::q0 | decider/copa/7e290dfd5a225933 | copa | decider | aps/super_glue | other (research) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | I came upon a puddle on the sidewalk. | false | What happened as a result? | [
{
"id": "I leapt over the puddle.",
"description": "I leapt over the puddle."
},
{
"id": "I played in the puddle.",
"description": "I played in the puddle."
}
] | choice | 0 | [
0.9999997846648723,
2.153351277748879e-7
] | null | null | null | {"task": "copa", "example": 218, "question": 0} | true |
decider/imdb/227::q0 | decider/imdb/db643d94e2830f72 | imdb | decider | stanfordnlp/imdb | other | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | This Italian film from the '70's is NOT even in the class with Dog Soldiers, The Howling, or even that awful American Werewolf in Paris, BUT...it is fun to watch. I'm talking about watching the lead actress, a stunning blonde, run amok in her birthday suit. We're talking about graphic, complete nudity...it's obvious th... | false | What is the sentiment of this movie review? | [
{
"id": "negative",
"description": "negative"
},
{
"id": "positive",
"description": "positive"
}
] | choice | 0 | [
0.5001219705608363,
0.4998780294391637
] | null | null | null | {"task": "imdb", "example": 227, "question": 0} | true |
decider/commands+fmt/263::q0 | decider/commands/4f845c0218ed29ba | commands | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | pip install --upgrade torch torchvision torchaudio | false | Does the command touch anything outside the current project directory (system files, other users, remote services, the whole disk)? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 0 | [
0.9999934964275259,
0.000006503572474176411
] | null | null | null | {"task": "commands+fmt", "example": 263, "question": 0} | true |
decider/commands+fmt/1398405::q0 | decider/commands/4f845c0218ed29ba | commands | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | pip install --upgrade torch torchvision torchaudio | false | How risky is running this shell command? | [
{
"id": "destructive",
"description": "destructive: irreversible: deletes or overwrites data, wipes disks or branches, force-pushes, changes credentials or exposes secrets"
},
{
"id": "safe",
"description": "safe: read-only, or creates files without touching existing ones; nothing to undo"
},
... | choice | 2 | [
0.0000047952969913209055,
0.000006028150098286229,
0.9999891765529104
] | null | null | null | {"task": "commands+fmt", "example": 1398405, "question": 0} | true |
decider/commands+fmt/1398405::q1 | decider/commands/4f845c0218ed29ba | commands | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | pip install --upgrade torch torchvision torchaudio | false | Does the command touch anything outside the current project directory (system files, other users, remote services, the whole disk)? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 0 | [
0.999992303557738,
0.000007696442261997626
] | null | null | null | {"task": "commands+fmt", "example": 1398405, "question": 1} | true |
decider/liar2/289::q0 | decider/liar2/b850663e791f1072 | liar2 | decider | chengxuphd/liar2 | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Statement: Sen. Johnny Isakson voted for a trade agreement, "with China costing Georgia 93,000 jobs.
Speaker: jim barksdale (Running for U.S. Senate in Georgia )
Subject: trade
Date: September 29, 2016 | false | How truthful is this political statement, as rated by fact-checkers? | [
{
"id": "0",
"description": "0: pants on fire (absurdly false)"
},
{
"id": "1",
"description": "1: false"
},
{
"id": "2",
"description": "2: barely true"
},
{
"id": "3",
"description": "3: half true"
},
{
"id": "4",
"description": "4: mostly true"
},
{
... | score | 3 | [
0.061255019059976684,
0.04476383116712651,
0.1419837585336856,
0.7289018102244692,
0.020342521461166222,
0.0027530595535758557
] | null | null | null | {"task": "liar2", "example": 289, "question": 0} | false |
decider/rules_boolean/290::q0 | decider/rules_boolean/40c1936a07cbb5f4 | rules_boolean | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | {"name": "Hana Ali", "age": 57, "years_experience": 20, "degree": "master", "skills": [], "salary": 72, "remote": true, "visa_required": true, "city": "Denver", "referred_by": "Ines Tanaka"} | false | Apply the policy: assign to Rui by default, assign to Dana when `visa_required` is false. | [
{
"id": "assign to Dana",
"description": "assign to Dana"
},
{
"id": "assign to Rui",
"description": "assign to Rui"
}
] | choice | 1 | [
2.879859295328871e-7,
0.9999997120140705
] | null | null | null | {"task": "rules_boolean", "example": 290, "question": 0} | true |
decider/rules_boolean/91295::q0 | decider/rules_boolean/40c1936a07cbb5f4 | rules_boolean | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | {"name": "Hana Ali", "age": 57, "years_experience": 20, "degree": "master", "skills": [], "salary": 72, "remote": true, "visa_required": true, "city": "Denver", "referred_by": "Ines Tanaka"} | false | Rule: assign to Rui when `visa_required` is false, else assign to Dana. | [
{
"id": "assign to Rui",
"description": "assign to Rui"
},
{
"id": "assign to Dana",
"description": "assign to Dana"
}
] | choice | 1 | [
7.977724472905951e-7,
0.9999992022275528
] | null | null | null | {"task": "rules_boolean", "example": 91295, "question": 0} | true |
decider/agenttraj/307::q0 | decider/agenttraj/e8326ccbcd84eed1 | agenttraj | decider | AgentGym/AgentTraj-L | unknown | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Environment: webshop.
Task:
WebShop [SEP] Instruction: [SEP] Find me home office furniture sets for dining room, living room with color: brown | beige, and item shape: runner, and size: 7 ft 9 in x 10 ft 8 in, and price lower than 80.00 dollars [SEP] Search
Recent actions: search[home office furniture sets dining room... | false | Which action should the agent take next? | [
{
"id": "click[7 ft 9 in x 10 ft 8 in]",
"description": "click[7 ft 9 in x 10 ft 8 in]"
},
{
"id": "search[home office furniture sets dining~33c89988",
"description": "search[home office furniture sets dining room living room color brown beige item shape runner size 7 ft 9 in x 10 ft 8 i"
},
... | choice | 0 | [
0.9998142890894466,
0.00004539239244453852,
0.00012499147706810138,
0.0000035165678092549594,
0.000011810473231566471
] | null | null | null | {"task": "agenttraj", "example": 307, "question": 0} | false |
decider/helpsteer3_pref+iso/329::q0 | decider/helpsteer3_pref/f301071a8edd8af7 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: Write a slogan for me something like this " The way it should be, Perfect"
Response 1:
Here are a few slogan suggestions that follow a similar structure:
1. "The way it was meant to be, Flawless"
2. "The way it ought to be, Exceptional"
3. "The way it was designed to be, Impeccable"
4. "The way i... | false | Which response is the better final reply, and by how much?
Proposed answer: both are about the same
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.00004373730763535546,
0.9999562626923646
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 329, "question": 0} | false |
decider/helpsteer3_pref/161548::q0 | decider/helpsteer3_pref/f301071a8edd8af7 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: Write a slogan for me something like this " The way it should be, Perfect"
Response 1:
Here are a few slogan suggestions that follow a similar structure:
1. "The way it was meant to be, Flawless"
2. "The way it ought to be, Exceptional"
3. "The way it was designed to be, Impeccable"
4. "The way i... | false | Which response is the better final reply, and by how much? | [
{
"id": "-3",
"description": "-3: response 1 is much better"
},
{
"id": "-2",
"description": "-2: response 1 is better"
},
{
"id": "-1",
"description": "-1: response 1 is slightly better"
},
{
"id": "0",
"description": "0: both are about the same"
},
{
"id": "1",
... | score | 0 | [
0.5000011443182915,
0.4999849950335439,
0.000007002163832733,
9.875605561350108e-7,
9.52072815665882e-8,
0.000005503916839471695,
2.7179965472781354e-7
] | null | null | null | {"task": "helpsteer3_pref", "example": 161548, "question": 0} | true |
decider/helpsteer3_pref+iso/365863::q0 | decider/helpsteer3_pref/f301071a8edd8af7 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: Write a slogan for me something like this " The way it should be, Perfect"
Response 1:
Here are a few slogan suggestions that follow a similar structure:
1. "The way it was meant to be, Flawless"
2. "The way it ought to be, Exceptional"
3. "The way it was designed to be, Impeccable"
4. "The way i... | false | Which response is the better final reply, and by how much?
Proposed answer: response 2 is much better
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.000011074004270439882,
0.9999889259957295
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 365863, "question": 0} | false |
decider/helpsteer3_pref+iso/645878::q0 | decider/helpsteer3_pref/f301071a8edd8af7 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: Write a slogan for me something like this " The way it should be, Perfect"
Response 1:
Here are a few slogan suggestions that follow a similar structure:
1. "The way it was meant to be, Flawless"
2. "The way it ought to be, Exceptional"
3. "The way it was designed to be, Impeccable"
4. "The way i... | false | Which response is the better final reply, and by how much?
Proposed answer: response 2 is slightly better
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.000004231879387044085,
0.999995768120613
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 645878, "question": 0} | true |
decider/helpsteer3_pref+iso/664295::q0 | decider/helpsteer3_pref/f301071a8edd8af7 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: Write a slogan for me something like this " The way it should be, Perfect"
Response 1:
Here are a few slogan suggestions that follow a similar structure:
1. "The way it was meant to be, Flawless"
2. "The way it ought to be, Exceptional"
3. "The way it was designed to be, Impeccable"
4. "The way i... | false | Which response is the better final reply, and by how much?
Proposed answer: response 1 is much better
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 0 | [
0.9999963801459141,
0.000003619854085946064
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 664295, "question": 0} | true |
decider/helpsteer3_pref+iso/839229::q0 | decider/helpsteer3_pref/f301071a8edd8af7 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: Write a slogan for me something like this " The way it should be, Perfect"
Response 1:
Here are a few slogan suggestions that follow a similar structure:
1. "The way it was meant to be, Flawless"
2. "The way it ought to be, Exceptional"
3. "The way it was designed to be, Impeccable"
4. "The way i... | false | Which response is the better final reply, and by how much?
Proposed answer: response 1 is slightly better
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.49999552141502446,
0.5000044785849755
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 839229, "question": 0} | false |
decider/helpsteer3_pref+iso/950351::q0 | decider/helpsteer3_pref/f301071a8edd8af7 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: Write a slogan for me something like this " The way it should be, Perfect"
Response 1:
Here are a few slogan suggestions that follow a similar structure:
1. "The way it was meant to be, Flawless"
2. "The way it ought to be, Exceptional"
3. "The way it was designed to be, Impeccable"
4. "The way i... | false | Which response is the better final reply, and by how much?
Proposed answer: response 1 is better
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.4955031520922353,
0.5044968479077647
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 950351, "question": 0} | false |
decider/helpsteer3_pref+iso/1228808::q0 | decider/helpsteer3_pref/f301071a8edd8af7 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: Write a slogan for me something like this " The way it should be, Perfect"
Response 1:
Here are a few slogan suggestions that follow a similar structure:
1. "The way it was meant to be, Flawless"
2. "The way it ought to be, Exceptional"
3. "The way it was designed to be, Impeccable"
4. "The way i... | false | Which response is the better final reply, and by how much?
Proposed answer: response 2 is better
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.000007008740504481901,
0.9999929912594955
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 1228808, "question": 0} | true |
decider/amazon_stars/351::q0 | decider/amazon_stars/08e8d439f67d55dc | amazon_stars | decider | SetFit/amazon_reviews_multi_en | apache-2.0 (mirror; original corpus had its own terms) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | My child loves garbage trucks so this was a perfect gift to give. It comes with two garbage bins and the back of the truck opens up. You can lift the bins with the arms and dump the garbage in the truck. This truck provides endless fun. Do be careful with the lift arms as it is pretty thin and it can break. Other than ... | false | How many stars did this reviewer most likely give? | [
{
"id": "2 stars",
"description": "2 stars"
},
{
"id": "5 stars",
"description": "5 stars"
},
{
"id": "1 star",
"description": "1 star"
},
{
"id": "4 stars",
"description": "4 stars"
},
{
"id": "3 stars",
"description": "3 stars"
}
] | choice | 1 | [
1.6850338591755177e-8,
0.5000001523183801,
0.0001333138764485707,
0.4998664518846535,
6.507017922776375e-8
] | null | null | null | {"task": "amazon_stars", "example": 351, "question": 0} | true |
decider/amazon_stars+fmt/106825::q0 | decider/amazon_stars/08e8d439f67d55dc | amazon_stars | decider | SetFit/amazon_reviews_multi_en | apache-2.0 (mirror; original corpus had its own terms) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | My child loves garbage trucks so this was a perfect gift to give. It comes with two garbage bins and the back of the truck opens up. You can lift the bins with the arms and dump the garbage in the truck. This truck provides endless fun. Do be careful with the lift arms as it is pretty thin and it can break. Other than ... | false | How many stars did this reviewer most likely give? | [
{
"id": "5 stars",
"description": "5 stars"
},
{
"id": "cooking recipe",
"description": "cooking recipe"
},
{
"id": "unable to verify identity",
"description": "unable to verify identity"
},
{
"id": "meaning of life",
"description": "meaning of life"
},
{
"id": "e... | choice | 0 | [
0.5001228831179995,
7.745257170066505e-7,
5.391109926140845e-7,
0.000002121613144057874,
3.085648654937224e-7,
0.0000012991061943938665,
0.0000012295053791879615,
9.396908017310435e-7,
9.576921036148085e-8,
5.455896301653464e-7,
1.2195420097697044e-7,
0.000005642309897603475,
1.3090096276414... | null | null | null | {"task": "amazon_stars+fmt", "example": 106825, "question": 0} | false |
decider/rules_threshold/369::q0 | decider/rules_threshold/bac105d5a21afe6a | rules_threshold | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Name: Diego Byrne
Age: 60
Years experience: 27
Degree: master
Skills: excel, go, figma
Salary: 131
Remote: no
Visa required: no
City: Lisbon
Referred by: (empty) | false | If the years experience is more than 15, choose assign to Dana; otherwise choose assign to Rui. | [
{
"id": "assign to Rui",
"description": "assign to Rui"
},
{
"id": "assign to Dana",
"description": "assign to Dana"
}
] | choice | 1 | [
7.072547575980253e-8,
0.9999999292745243
] | null | null | null | {"task": "rules_threshold", "example": 369, "question": 0} | true |
decider/rules_threshold/492799::q0 | decider/rules_threshold/bac105d5a21afe6a | rules_threshold | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Name: Diego Byrne
Age: 60
Years experience: 27
Degree: master
Skills: excel, go, figma
Salary: 131
Remote: no
Visa required: no
City: Lisbon
Referred by: (empty) | false | Apply the policy: assign to Dana by default, assign to Rui when the years experience is more than 15. | [
{
"id": "assign to Dana",
"description": "assign to Dana"
},
{
"id": "assign to Rui",
"description": "assign to Rui"
}
] | choice | 1 | [
0.000003508378259817804,
0.9999964916217402
] | null | null | null | {"task": "rules_threshold", "example": 492799, "question": 0} | false |
decider/mmlu/403::q0 | decider/mmlu/8b621106c9b5af68 | mmlu | decider | cais/mmlu | mit | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Saturday, March 20, 2010 Dear Mom and Dad, The time has come: I'm graduating, and I thank you both for all your help during these 15 years. I still remember when you used to say, "You can be whatever you want". I'm so glad you are my parents, especially when I hear kids say they don't get on well with theirs. I'm glad ... | false | Which option is correct? | [
{
"id": "help",
"description": "help"
},
{
"id": "thanks",
"description": "thanks"
},
{
"id": "support",
"description": "support"
},
{
"id": "wishes",
"description": "wishes"
}
] | choice | 1 | [
1.0786088989743477e-8,
0.9999999159075739,
1.294121449263545e-8,
6.03651226277212e-8
] | null | null | null | {"task": "mmlu", "example": 403, "question": 0} | true |
decider/prosocial_safety+iso/460::q0 | decider/prosocial_safety/50484ecd56752fe0 | prosocial_safety | decider | allenai/prosocial-dialog | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Because I want her to know there are still rules in our relationship. | false | How much does this utterance call for a cautious or corrective response?
Proposed answer: needs intervention
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.0003327917720643519,
0.9996672082279356
] | null | null | null | {"task": "prosocial_safety+iso", "example": 460, "question": 0} | false |
decider/prosocial_safety+iso/159594::q0 | decider/prosocial_safety/50484ecd56752fe0 | prosocial_safety | decider | allenai/prosocial-dialog | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Because I want her to know there are still rules in our relationship. | false | How much does this utterance call for a cautious or corrective response?
Proposed answer: casual, no safety concern
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.4999833440889932,
0.5000166559110069
] | null | null | null | {"task": "prosocial_safety+iso", "example": 159594, "question": 0} | false |
decider/prosocial_safety/443315::q0 | decider/prosocial_safety/50484ecd56752fe0 | prosocial_safety | decider | allenai/prosocial-dialog | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Because I want her to know there are still rules in our relationship. | false | How much does this utterance call for a cautious or corrective response? | [
{
"id": "0",
"description": "0: casual, no safety concern"
},
{
"id": "1",
"description": "1: possibly needs caution"
},
{
"id": "2",
"description": "2: probably needs caution"
},
{
"id": "3",
"description": "3: needs caution"
},
{
"id": "4",
"description": "4... | score | 3 | [
6.511512278309084e-7,
0.25025786375969195,
0.24973935649395382,
0.5000019758142591,
1.5278086730901934e-7
] | null | null | null | {"task": "prosocial_safety", "example": 443315, "question": 0} | true |
decider/prosocial_safety+iso/949043::q0 | decider/prosocial_safety/50484ecd56752fe0 | prosocial_safety | decider | allenai/prosocial-dialog | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Because I want her to know there are still rules in our relationship. | false | How much does this utterance call for a cautious or corrective response?
Proposed answer: possibly needs caution
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.49999176051982,
0.50000823948018
] | null | null | null | {"task": "prosocial_safety+iso", "example": 949043, "question": 0} | false |
decider/prosocial_safety+iso/1144780::q0 | decider/prosocial_safety/50484ecd56752fe0 | prosocial_safety | decider | allenai/prosocial-dialog | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Because I want her to know there are still rules in our relationship. | false | How much does this utterance call for a cautious or corrective response?
Proposed answer: probably needs caution
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.4999370120667067,
0.5000629879332933
] | null | null | null | {"task": "prosocial_safety+iso", "example": 1144780, "question": 0} | true |
decider/prosocial_safety+iso/1478913::q0 | decider/prosocial_safety/50484ecd56752fe0 | prosocial_safety | decider | allenai/prosocial-dialog | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Because I want her to know there are still rules in our relationship. | false | How much does this utterance call for a cautious or corrective response?
Proposed answer: needs caution
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 0 | [
0.9999591946505936,
0.00004080534940642879
] | null | null | null | {"task": "prosocial_safety+iso", "example": 1478913, "question": 0} | true |
decider/hellaswag/462::q0 | decider/hellaswag/0246fa86e16ba768 | hellaswag | decider | Rowan/hellaswag | unknown | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | [header] How to make quick stuffed peppers (turkey ) [title] Wash your hands. [title] Prepare your work space. [step] Make sure your counters are clean and that you've got enough room to move around, and have all of your utensils out and clean. | false | Which ending is the most plausible continuation? | [
{
"id": "[title] Slice the peppers lengthwise and~d0f41a45",
"description": "[title] Slice the peppers lengthwise and remove any blackheads from each stem. [step] Slice one half diagonally, slightly shorter than the other half."
},
{
"id": "[title] Pre-heat the oven to 400 degrees~93e14672",
"de... | choice | 1 | [
4.700727376976857e-9,
0.9999999111240423,
2.8649091462657952e-8,
5.5526138937497404e-8
] | null | null | null | {"task": "hellaswag", "example": 462, "question": 0} | true |
decider/insincere_questions/491::q0 | decider/insincere_questions/e27560e39c1ae5bf | insincere_questions | decider | SetFit/insincere-questions | unknown | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | What if Benito Mussolini never rose to power? | false | Is this forum question insincere (rhetorical, provocative, or not seeking a real answer)? | [
{
"id": "sincere",
"description": "sincere"
},
{
"id": "insincere",
"description": "insincere"
}
] | choice | 0 | [
0.9999987599414082,
0.0000012400585918517389
] | null | null | null | {"task": "insincere_questions", "example": 491, "question": 0} | false |
decider/amazon_stars+iso/492::q0 | decider/amazon_stars/eae695cc10d1de55 | amazon_stars | decider | SetFit/amazon_reviews_multi_en | apache-2.0 (mirror; original corpus had its own terms) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | This product will rust within one month of use in the cooler. It fits the new style RTIC cooler just fine but will rust. The plastic coating around the wire rack will crack thus causing rust. Not worth the money... | false | How many stars did this reviewer most likely give?
Proposed answer: 4 stars
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.000016153042286579758,
0.9999838469577135
] | null | null | null | {"task": "amazon_stars+iso", "example": 492, "question": 0} | true |
decider/amazon_stars+iso/48244::q0 | decider/amazon_stars/eae695cc10d1de55 | amazon_stars | decider | SetFit/amazon_reviews_multi_en | apache-2.0 (mirror; original corpus had its own terms) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | This product will rust within one month of use in the cooler. It fits the new style RTIC cooler just fine but will rust. The plastic coating around the wire rack will crack thus causing rust. Not worth the money... | false | How many stars did this reviewer most likely give?
Proposed answer: 3 stars
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.000003969292209961579,
0.9999960307077901
] | null | null | null | {"task": "amazon_stars+iso", "example": 48244, "question": 0} | true |
decider/amazon_stars/180765::q0 | decider/amazon_stars/eae695cc10d1de55 | amazon_stars | decider | SetFit/amazon_reviews_multi_en | apache-2.0 (mirror; original corpus had its own terms) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | This product will rust within one month of use in the cooler. It fits the new style RTIC cooler just fine but will rust. The plastic coating around the wire rack will crack thus causing rust. Not worth the money... | false | How many stars did this reviewer most likely give? | [
{
"id": "4 stars",
"description": "4 stars"
},
{
"id": "3 stars",
"description": "3 stars"
},
{
"id": "1 star",
"description": "1 star"
},
{
"id": "5 stars",
"description": "5 stars"
},
{
"id": "2 stars",
"description": "2 stars"
}
] | choice | 2 | [
8.444046677584947e-8,
9.568402921511315e-8,
0.9999971637653213,
0.0000025031595086612876,
1.5295067399338095e-7
] | null | null | null | {"task": "amazon_stars", "example": 180765, "question": 0} | true |
decider/amazon_stars+iso/1369251::q0 | decider/amazon_stars/eae695cc10d1de55 | amazon_stars | decider | SetFit/amazon_reviews_multi_en | apache-2.0 (mirror; original corpus had its own terms) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | This product will rust within one month of use in the cooler. It fits the new style RTIC cooler just fine but will rust. The plastic coating around the wire rack will crack thus causing rust. Not worth the money... | false | How many stars did this reviewer most likely give?
Proposed answer: 5 stars
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.0000017373448146635555,
0.9999982626551853
] | null | null | null | {"task": "amazon_stars+iso", "example": 1369251, "question": 0} | false |
decider/amazon_stars+iso/1415883::q0 | decider/amazon_stars/eae695cc10d1de55 | amazon_stars | decider | SetFit/amazon_reviews_multi_en | apache-2.0 (mirror; original corpus had its own terms) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | This product will rust within one month of use in the cooler. It fits the new style RTIC cooler just fine but will rust. The plastic coating around the wire rack will crack thus causing rust. Not worth the money... | false | How many stars did this reviewer most likely give?
Proposed answer: 1 star
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 0 | [
0.9999999622229823,
3.77770177079649e-8
] | null | null | null | {"task": "amazon_stars+iso", "example": 1415883, "question": 0} | false |
decider/amazon_stars+iso/1444229::q0 | decider/amazon_stars/eae695cc10d1de55 | amazon_stars | decider | SetFit/amazon_reviews_multi_en | apache-2.0 (mirror; original corpus had its own terms) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | This product will rust within one month of use in the cooler. It fits the new style RTIC cooler just fine but will rust. The plastic coating around the wire rack will crack thus causing rust. Not worth the money... | false | How many stars did this reviewer most likely give?
Proposed answer: 2 stars
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.000010579647678822068,
0.9999894203523212
] | null | null | null | {"task": "amazon_stars+iso", "example": 1444229, "question": 0} | false |
decider/go_emotions/556::q0 | decider/go_emotions/dafa3412539ee426 | go_emotions | decider | google-research-datasets/go_emotions | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | I REALLY love martini glasses. But not this much. | false | Which emotion best describes this comment? | [
{
"id": "anger",
"description": "anger"
},
{
"id": "grief",
"description": "grief"
},
{
"id": "gratitude",
"description": "gratitude"
},
{
"id": "disapproval",
"description": "disapproval"
},
{
"id": "pride",
"description": "pride"
},
{
"id": "sadness"... | choice | 8 | [
3.036751047752697e-8,
4.452952914304562e-7,
3.8527860065939445e-8,
0.4999938031116242,
9.742627824829693e-7,
0.0000029313608732617843,
7.506042824125234e-8,
2.4884563638132333e-7,
0.500000152384357,
0.000001300783636383648
] | null | null | null | {"task": "go_emotions", "example": 556, "question": 0} | true |
decider/massive_intent/571::q0 | decider/massive_intent/a6a6fa49f9aa4d4c | massive_intent | decider | SetFit/amazon_massive_intent_en-US | unknown | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | start radio station for me | false | What is the intent of this voice-assistant utterance? | [
{
"id": "audio volume other",
"description": "audio volume other"
},
{
"id": "social post",
"description": "social post"
},
{
"id": "takeaway query",
"description": "takeaway query"
},
{
"id": "iot hue lightup",
"description": "iot hue lightup"
},
{
"id": "alarm r... | choice | 8 | [
1.2903643551094847e-7,
9.349566071508836e-8,
1.2440503146082607e-7,
4.5008214709084034e-8,
2.775808465220507e-8,
7.89907979600955e-8,
2.2646581385176264e-7,
0.0000010335519474425216,
0.9999981923439191,
4.894409466628431e-8
] | null | null | null | {"task": "massive_intent", "example": 571, "question": 0} | true |
decider/massive_intent+fmt/1116439::q0 | decider/massive_intent/a6a6fa49f9aa4d4c | massive_intent | decider | SetFit/amazon_massive_intent_en-US | unknown | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | start radio station for me | false | What is the intent of this voice-assistant utterance? | [
{
"id": "lists query",
"description": "lists query"
},
{
"id": "takeaway query",
"description": "takeaway query"
},
{
"id": "calendar set",
"description": "calendar set"
},
{
"id": "social query",
"description": "social query"
},
{
"id": "play game",
"descript... | choice | 11 | [
1.2513092136060568e-7,
3.54568978359814e-8,
1.906881440514755e-8,
2.4970834301475786e-7,
1.5920567128156595e-7,
9.682343646472076e-9,
1.8154997609448238e-8,
2.9534456671715343e-8,
3.000657463618943e-8,
1.5259921690067291e-7,
3.7512605522600386e-8,
0.9999987004223051,
3.122759670941799e-7,
... | null | null | null | {"task": "massive_intent+fmt", "example": 1116439, "question": 0} | true |
decider/tweet_sentiment/639::q0 | decider/tweet_sentiment/2032b8afc8935ce1 | tweet_sentiment | decider | cardiffnlp/tweet_eval | unknown | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | @user I knew u in 8th grade and I'm pretty sure it probably had something to do with twilight | false | What is the sentiment of this tweet? | [
{
"id": "positive",
"description": "positive"
},
{
"id": "neutral",
"description": "neutral"
},
{
"id": "negative",
"description": "negative"
}
] | choice | 1 | [
7.0701160260655e-7,
0.9999992743989399,
1.8589457572546645e-8
] | null | null | null | {"task": "tweet_sentiment", "example": 639, "question": 0} | false |
decider/dbpedia/660::q0 | decider/dbpedia/d282ac33d04f17ad | dbpedia | decider | fancyzhx/dbpedia_14 | cc-by-sa-3.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | North River Mountain: North River Mountain is a mountain ridge that runs southwest to northeast in Hampshire County West Virginia United States. Named for the North River that parallels its western flanks for its entire length North River Mountain reaches its highest elevation of 2149 feet (655 m) near the community o... | false | What kind of entity does this encyclopedia text describe? | [
{
"id": "Animal",
"description": "Animal"
},
{
"id": "Company",
"description": "Company"
},
{
"id": "Athlete",
"description": "Athlete"
},
{
"id": "NaturalPlace",
"description": "NaturalPlace"
},
{
"id": "Film",
"description": "Film"
},
{
"id": "Office... | choice | 3 | [
3.774176298273054e-9,
1.9686430776325413e-8,
2.648111602458443e-9,
0.9999995897453391,
3.306764406422275e-8,
2.305176551579396e-9,
1.2493895348635266e-8,
4.5197913149764816e-8,
1.079668092440978e-7,
1.8311450369199615e-7
] | null | null | null | {"task": "dbpedia", "example": 660, "question": 0} | true |
decider/ultrafeedback_pref/667::q0 | decider/ultrafeedback_pref/a5fa15e5cbb11057 | ultrafeedback_pref | decider | HuggingFaceH4/ultrafeedback_binarized | mit | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Prompt:
Assume we have a band-limited signal and an ideal sampling function (Dirac Comb) with frequency twice that of the band-limited signal. We know our signal is band-limited because we filtered it with an ideal brick-wall filter with a pass-band equal to the band-width of the signal; thus the signal is strictly ban... | false | Which response is the better answer to the prompt? | [
{
"id": "response A",
"description": "response A"
},
{
"id": "response B",
"description": "response B"
}
] | choice | 1 | [
0.0010755279618728112,
0.9989244720381272
] | null | null | null | {"task": "ultrafeedback_pref", "example": 667, "question": 0} | true |
decider/sms_spam/672::q0 | decider/sms_spam/5543e148cb18d353 | sms_spam | decider | ucirvine/sms_spam | unknown | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Thanks for your Ringtone Order, Reference T91. You will be charged GBP 4 per week. You can unsubscribe at anytime by calling customer services on 09057039994 | false | Is this SMS message spam? | [
{
"id": "spam",
"description": "spam"
},
{
"id": "ham (legitimate)",
"description": "ham (legitimate)"
}
] | choice | 0 | [
0.7500717544069282,
0.2499282455930718
] | null | null | null | {"task": "sms_spam", "example": 672, "question": 0} | true |
decider/snli/687::q0 | decider/snli/077615132f2b245a | snli | decider | stanfordnlp/snli | cc-by-sa-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Text 1: Men build castle-like sculptures out of dark sand by the shore.
Text 2: Men build something out of sand. | false | What is the logical relation between text 1 and text 2? | [
{
"id": "entailment (text 2 follows from text 1)",
"description": "entailment (text 2 follows from text 1)"
},
{
"id": "contradiction",
"description": "contradiction"
},
{
"id": "neutral",
"description": "neutral"
}
] | choice | 0 | [
0.9999994234980139,
4.960374124730255e-7,
8.046457364640254e-8
] | null | null | null | {"task": "snli", "example": 687, "question": 0} | false |
decider/agenttraj/691::q0 | decider/agenttraj/a5aadb4daf2e1ca6 | agenttraj | decider | AgentGym/AgentTraj-L | unknown | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Environment: sciworld.
Task:
Your task is to determine if unknown substance G is electrically conductive. The unknown substance G is located around the workshop. First, focus on the unknown substance G. If it is electrically conductive, place it in the orange box. If it is electrically nonconductive, place it in the ye... | false | Which action should the agent take next? | [
{
"id": "go to bedroom",
"description": "go to bedroom"
},
{
"id": "open door to living room",
"description": "open door to living room"
},
{
"id": "move gallium to metal pot containing gallium",
"description": "move gallium to metal pot containing gallium"
},
{
"id": "pick u... | choice | 0 | [
0.9999870514576441,
0.0000036562145831946997,
0.0000029458341813784162,
0.0000060952628243790554,
2.5123076698876896e-7
] | null | null | null | {"task": "agenttraj", "example": 691, "question": 0} | false |
decider/race/709::q0 | decider/race/6c13bbf9ab20c96a | race | decider | ehovy/race | other (non-commercial research purposes only) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Article: Dear SJ,
Losing a best friend is never easy.
Your problem, is not just that you miss your best friend, it is that you feel empty and lost without her friendship.
It takes time to get over a lost, and during that time, your mind is getting used to _ . This is usually a good thing, even if it feels like a bad th... | false | Which option correctly answers the question about the article? | [
{
"id": "feel more independent and confident",
"description": "feel more independent and confident"
},
{
"id": "comfort her former friend",
"description": "comfort her former friend"
},
{
"id": "take revenge on her former friend",
"description": "take revenge on her former friend"
... | choice | 0 | [
0.9999996210069877,
2.52666396888849e-7,
1.5262694128038295e-8,
1.1106392131812866e-7
] | null | null | null | {"task": "race", "example": 709, "question": 0} | true |
decider/medqa/722::q0 | decider/medqa/4ae28e982ae76a1f | medqa | decider | GBaker/MedQA-USMLE-4-options | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | A 45-year-old man is brought into the clinic by his wife. She reports that her husband has been feeling down since he lost a big project at work 2 months ago. The patient says he feels unmotivated to work or do things around the house. He also says he is not eating or sleeping as usual and spends most of his day pacing... | false | Which option is the correct answer? | [
{
"id": "Sleepwalking",
"description": "Sleepwalking"
},
{
"id": "Respiratory depression",
"description": "Respiratory depression"
},
{
"id": "Dry mouth",
"description": "Dry mouth"
},
{
"id": "Lightheadedness",
"description": "Lightheadedness"
}
] | choice | 3 | [
9.085549674397379e-7,
5.637744658256906e-8,
6.576232647197766e-8,
0.9999989693052596
] | null | null | null | {"task": "medqa", "example": 722, "question": 0} | false |
decider/multirc/747::q0 | decider/multirc/600c2bdf98a76436 | multirc | decider | aps/super_glue | other (research) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Paragraph: Comanche Indians have trapped the evil of their land in a monster made up of a mixture of other animals . The Comanche have placed a protective spear in the ground to contain the evil . A young boy pulls the spear from the ground while pocketing a rabbit foot talisman . He starts to play with the spear . ... | false | Is the candidate answer correct according to the paragraph? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.0000014320359272718434,
0.9999985679640727
] | null | null | null | {"task": "multirc", "example": 747, "question": 0} | false |
decider/multirc+fmt/1278777::q0 | decider/multirc/600c2bdf98a76436 | multirc | decider | aps/super_glue | other (research) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Paragraph: Comanche Indians have trapped the evil of their land in a monster made up of a mixture of other animals . The Comanche have placed a protective spear in the ground to contain the evil . A young boy pulls the spear from the ground while pocketing a rabbit foot talisman . He starts to play with the spear . ... | false | Is the candidate answer correct according to the paragraph? | [
{
"id": "yes",
"description": "yes: Inputs where the candidate answer is factually supported by the provided paragraph text. (not: Inputs where the answer contradicts the text or cannot be verified from it.)"
},
{
"id": "no",
"description": "no: Inputs where the candidate answer is factually inc... | noul | 1 | [
0.0000014904297997884394,
0.9999985095702002
] | null | null | null | {"task": "multirc+fmt", "example": 1278777, "question": 0} | false |
decider/helpsteer2+fmt/751::q0 | decider/helpsteer2/9cbf01245eddeb81 | helpsteer2 | decider | nvidia/HelpSteer2 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Prompt:
In the Shopify Admin API, I am trying to make a list of order counts by day and I would like to use the `orders/count` endpoint provided but am worrying about hitting the rate limit for the API. Do you know what the cost of this query is and how many you could run in a single second without getting rate limited... | false | How helpful is the response to the prompt? | [
{
"id": "0",
"description": "0: not helpful"
},
{
"id": "1",
"description": "1: slightly helpful"
},
{
"id": "2",
"description": "2: partially helpful"
},
{
"id": "3",
"description": "3: mostly helpful"
},
{
"id": "4",
"description": "4: extremely helpful"
}... | score | 3 | [
1.926790905990342e-7,
0.4999994880849427,
2.8630859535683636e-7,
0.5000000257737471,
7.153624227988832e-9
] | null | null | null | {"task": "helpsteer2+fmt", "example": 751, "question": 0} | false |
decider/helpsteer2/624074::q0 | decider/helpsteer2/9cbf01245eddeb81 | helpsteer2 | decider | nvidia/HelpSteer2 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Prompt:
In the Shopify Admin API, I am trying to make a list of order counts by day and I would like to use the `orders/count` endpoint provided but am worrying about hitting the rate limit for the API. Do you know what the cost of this query is and how many you could run in a single second without getting rate limited... | false | How helpful is the response to the prompt? | [
{
"id": "0",
"description": "0: not helpful"
},
{
"id": "1",
"description": "1: slightly helpful"
},
{
"id": "2",
"description": "2: partially helpful"
},
{
"id": "3",
"description": "3: mostly helpful"
},
{
"id": "4",
"description": "4: extremely helpful"
}... | score | 3 | [
0.0002549562509371833,
0.49970339020959875,
0.00003861993690418277,
0.5000021264169895,
9.07185570413842e-7
] | null | null | null | {"task": "helpsteer2", "example": 624074, "question": 0} | false |
decider/helpsteer2/624074::q1 | decider/helpsteer2/9cbf01245eddeb81 | helpsteer2 | decider | nvidia/HelpSteer2 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Prompt:
In the Shopify Admin API, I am trying to make a list of order counts by day and I would like to use the `orders/count` endpoint provided but am worrying about hitting the rate limit for the API. Do you know what the cost of this query is and how many you could run in a single second without getting rate limited... | false | How correct and accurate is the response? | [
{
"id": "0",
"description": "0: completely incorrect"
},
{
"id": "1",
"description": "1: mostly incorrect"
},
{
"id": "2",
"description": "2: partially correct"
},
{
"id": "3",
"description": "3: mostly correct"
},
{
"id": "4",
"description": "4: completely co... | score | 3 | [
0.00020152767646245415,
0.24378282970165158,
0.2557059217931701,
0.5002555677442794,
0.000054153084436516537
] | null | null | null | {"task": "helpsteer2", "example": 624074, "question": 1} | false |
decider/helpsteer2/624074::q2 | decider/helpsteer2/9cbf01245eddeb81 | helpsteer2 | decider | nvidia/HelpSteer2 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Prompt:
In the Shopify Admin API, I am trying to make a list of order counts by day and I would like to use the `orders/count` endpoint provided but am worrying about hitting the rate limit for the API. Do you know what the cost of this query is and how many you could run in a single second without getting rate limited... | false | How coherent and clear is the response? | [
{
"id": "0",
"description": "0: incoherent"
},
{
"id": "1",
"description": "1: mostly incoherent"
},
{
"id": "2",
"description": "2: partially coherent"
},
{
"id": "3",
"description": "3: mostly coherent"
},
{
"id": "4",
"description": "4: perfectly coherent"
... | score | 4 | [
4.5147807564718026e-7,
0.000005774859121582082,
0.000041843544352073236,
0.49994969098069997,
0.5000022391377508
] | null | null | null | {"task": "helpsteer2", "example": 624074, "question": 2} | false |
decider/helpsteer2/624074::q3 | decider/helpsteer2/9cbf01245eddeb81 | helpsteer2 | decider | nvidia/HelpSteer2 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Prompt:
In the Shopify Admin API, I am trying to make a list of order counts by day and I would like to use the `orders/count` endpoint provided but am worrying about hitting the rate limit for the API. Do you know what the cost of this query is and how many you could run in a single second without getting rate limited... | false | What level of expertise does the response's content require? | [
{
"id": "0",
"description": "0: basic, anyone could write it"
},
{
"id": "1",
"description": "1: simple"
},
{
"id": "2",
"description": "2: intermediate"
},
{
"id": "3",
"description": "3: advanced"
},
{
"id": "4",
"description": "4: expert-level"
}
] | score | 2 | [
0.000010944137392627275,
0.25000188354994385,
0.7499714550531554,
0.000015296415338187217,
4.208441699685381e-7
] | null | null | null | {"task": "helpsteer2", "example": 624074, "question": 3} | false |
decider/helpsteer2/624074::q4 | decider/helpsteer2/9cbf01245eddeb81 | helpsteer2 | decider | nvidia/HelpSteer2 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Prompt:
In the Shopify Admin API, I am trying to make a list of order counts by day and I would like to use the `orders/count` endpoint provided but am worrying about hitting the rate limit for the API. Do you know what the cost of this query is and how many you could run in a single second without getting rate limited... | false | How verbose is the response relative to what the prompt asks for? | [
{
"id": "0",
"description": "0: far too short"
},
{
"id": "1",
"description": "1: somewhat short"
},
{
"id": "2",
"description": "2: adequate length"
},
{
"id": "3",
"description": "3: somewhat long"
},
{
"id": "4",
"description": "4: far too long"
}
] | score | 3 | [
0.000007991950262691938,
0.0000012733247469298606,
0.0000013844224980086824,
0.5010273416061863,
0.498962008696306
] | null | null | null | {"task": "helpsteer2", "example": 624074, "question": 4} | false |
decider/rules_threshold/809::q0 | decider/rules_threshold/ecd73122fa7e5493 | rules_threshold | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | {"subject": "Refund request", "priority": "low", "category": "technical", "assignee": "Sofia Haddad", "age_hours": 175, "tier": "basic", "resolved": false, "replies": 7, "reporter": "Hana Costa"} | false | Rule: assign to Dana when the replies is below 8, else assign to Rui. | [
{
"id": "assign to Dana",
"description": "assign to Dana"
},
{
"id": "assign to Rui",
"description": "assign to Rui"
}
] | choice | 0 | [
0.9999968907261497,
0.0000031092738502875133
] | null | null | null | {"task": "rules_threshold", "example": 809, "question": 0} | true |
decider/commands+fmt/891::q0 | decider/commands/013c4b3e57aaf2af | commands | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | dig @8.8.8.8 example.com A +short | false | How risky is running this shell command? | [
{
"id": "caution",
"description": "caution: changes files, packages or settings but can be undone or redone"
},
{
"id": "safe",
"description": "safe: read-only, or creates files without touching existing ones; nothing to undo"
},
{
"id": "destructive",
"description": "destructive: ir... | choice | 1 | [
0.0000011420876031532194,
0.999982544577969,
0.00001631333442787619
] | null | null | null | {"task": "commands+fmt", "example": 891, "question": 0} | false |
decider/commands+fmt/891::q1 | decider/commands/013c4b3e57aaf2af | commands | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | dig @8.8.8.8 example.com A +short | false | Does the command touch anything outside the current project directory (system files, other users, remote services, the whole disk)? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 0 | [
0.9999964916462881,
0.000003508353711891296
] | null | null | null | {"task": "commands+fmt", "example": 891, "question": 1} | false |
decider/commands+fmt/505329::q0 | decider/commands/013c4b3e57aaf2af | commands | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | dig @8.8.8.8 example.com A +short | false | How risky is running this shell command? | [
{
"id": "destructive",
"description": "destructive: irreversible: deletes or overwrites data, wipes disks or branches, force-pushes, changes credentials or exposes secrets"
},
{
"id": "safe",
"description": "safe: read-only, or creates files without touching existing ones; nothing to undo"
},
... | choice | 1 | [
5.729806098045044e-7,
0.9999877094465592,
0.000011717572830812911
] | null | null | null | {"task": "commands+fmt", "example": 505329, "question": 0} | false |
decider/commands+fmt/505329::q1 | decider/commands/013c4b3e57aaf2af | commands | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | dig @8.8.8.8 example.com A +short | false | Does the command touch anything outside the current project directory (system files, other users, remote services, the whole disk)? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 0 | [
0.9999962653763556,
0.0000037346236444437635
] | null | null | null | {"task": "commands+fmt", "example": 505329, "question": 1} | true |
decider/commands+fmt/969371::q0 | decider/commands/013c4b3e57aaf2af | commands | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | dig @8.8.8.8 example.com A +short | false | Does the command touch anything outside the current project directory (system files, other users, remote services, the whole disk)? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 0 | [
0.9999752133208908,
0.000024786679109257272
] | null | null | null | {"task": "commands+fmt", "example": 969371, "question": 0} | true |
decider/commands+fmt/1011776::q0 | decider/commands/013c4b3e57aaf2af | commands | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | dig @8.8.8.8 example.com A +short | false | Does the command touch anything outside the current project directory (system files, other users, remote services, the whole disk)? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 0 | [
0.9997711125849523,
0.0002288874150477011
] | null | null | null | {"task": "commands+fmt", "example": 1011776, "question": 0} | true |
decider/toolace/913::q0 | decider/toolace/0da0c9b5f740c40f | toolace | decider | Team-ACE/ToolACE | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Available functions:
- game_score.submit: Submit a new high score for a specific retro game along with player details.
- identifySpeechEmotion: Identify the emotion expressed in a speech
- Get Category: Retrieve a specific category by its ID, belonging to a team.
- speakerIdentification: Identify the speaker from a giv... | false | Which function should be called to handle this request? | [
{
"id": "Get Category",
"description": "Get Category"
},
{
"id": "Get Handball Matches by Category and Date",
"description": "Get Handball Matches by Category and Date"
},
{
"id": "no function call needed",
"description": "no function call needed"
},
{
"id": "generateSpeech",... | choice | 3 | [
1.0193734706055492e-7,
9.801337268136785e-8,
9.562425275450496e-8,
0.999995786830202,
1.4368476922171894e-7,
1.818488213174355e-7,
0.0000030280329510626832,
5.64028283943053e-7
] | null | null | null | {"task": "toolace", "example": 913, "question": 0} | false |
decider/rules_form/916::q0 | decider/rules_form/b4ccf1414646a657 | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Northwind Clinic - New Patient Registration
ELEMENT Pane "Sidebar" value="" | false | What should be done with this element? | [
{
"id": "fill Mobile (old): +1 202 555 0241",
"description": "fill Mobile (old): +1 202 555 0241"
},
{
"id": "fill Town: Madrid",
"description": "fill Town: Madrid"
},
{
"id": "check",
"description": "check"
},
{
"id": "fill Account: ACCT-YDKHQGAO",
"description": "fill A... | choice | 4 | [
2.2492588627028966e-8,
4.4987342253292273e-8,
2.123506491355108e-7,
2.410997627397174e-7,
0.9999983871456152,
1.7792829601582334e-7,
3.519217689961507e-7,
1.6995900304536198e-7,
3.921149740317166e-7
] | null | null | null | {"task": "rules_form", "example": 916, "question": 0} | true |
decider/rules_form/4792::q0 | decider/rules_form/b4ccf1414646a657 | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Northwind Clinic - New Patient Registration
ELEMENT Pane "Sidebar" value="" | false | What should be done with this element? | [
{
"id": "fill Passport: P-YNMLHPS",
"description": "fill Passport: P-YNMLHPS"
},
{
"id": "fill Policy no: POL-0Q3UF6",
"description": "fill Policy no: POL-0Q3UF6"
},
{
"id": "fill Phone: +1 202 555 0280",
"description": "fill Phone: +1 202 555 0280"
},
{
"id": "click",
"d... | choice | 4 | [
9.552475230668614e-10,
3.278491966492859e-9,
3.469059121975086e-9,
3.5375626610794876e-7,
0.9999993972140456,
5.289168734091883e-9,
2.205296671702688e-9,
1.6292332001417643e-7,
3.4071767567677895e-9,
2.97100462821992e-9,
6.453092282288346e-8
] | null | null | null | {"task": "rules_form", "example": 4792, "question": 0} | true |
decider/rules_form/244214::q0 | decider/rules_form/b4ccf1414646a657 | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Northwind Clinic - New Patient Registration
ELEMENT Pane "Sidebar" value="" | false | Fill an empty field with its document value (also when the value is stale), check a required box that is unchecked, click the submit button; skip anything else. What applies here? | [
{
"id": "check",
"description": "check"
},
{
"id": "click",
"description": "click"
},
{
"id": "skip",
"description": "skip"
},
{
"id": "fill Account: ACCT-W6R6HKG5",
"description": "fill Account: ACCT-W6R6HKG5"
},
{
"id": "fill Notes: Nairobi",
"description": ... | choice | 2 | [
4.3607672431596576e-8,
1.5619985076584582e-8,
0.9999997876698875,
2.7298257320501094e-8,
3.2767691642047076e-9,
7.270005017983233e-8,
1.1886596844078082e-8,
3.794078145664351e-8
] | null | null | null | {"task": "rules_form", "example": 244214, "question": 0} | true |
decider/rules_form/373435::q0 | decider/rules_form/b4ccf1414646a657 | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Northwind Clinic - New Patient Registration
ELEMENT Pane "Sidebar" value="" | false | Fill an empty field with its document value (also when the value is stale), check a required box that is unchecked, click the submit button; skip anything else. What applies here? | [
{
"id": "fill Contact email: zara.novak@example.org",
"description": "fill Contact email: zara.novak@example.org"
},
{
"id": "fill Applicant: Sofia Ali",
"description": "fill Applicant: Sofia Ali"
},
{
"id": "fill Notes: Perth",
"description": "fill Notes: Perth"
},
{
"id": "... | choice | 5 | [
2.4527495129781082e-8,
9.847172681734617e-8,
1.9102044200271397e-8,
4.339272338146706e-8,
1.2541773415917153e-7,
0.9999995597583609,
1.0823274801242487e-8,
3.781407219801597e-8,
2.7131265381699766e-8,
2.971697974538676e-8,
1.2333180534015645e-8,
1.151114264191132e-8
] | null | null | null | {"task": "rules_form", "example": 373435, "question": 0} | true |
decider/rules_form/494798::q0 | decider/rules_form/b4ccf1414646a657 | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Northwind Clinic - New Patient Registration
ELEMENT Pane "Sidebar" value="" | false | What should the form-filling agent do with this element? | [
{
"id": "fill Employer: Example Trading",
"description": "fill Employer: Example Trading"
},
{
"id": "fill Contact email: sofia.fischer@example.org",
"description": "fill Contact email: sofia.fischer@example.org"
},
{
"id": "check",
"description": "check"
},
{
"id": "fill Pas... | choice | 7 | [
3.3351724257341906e-7,
2.6383301900723604e-7,
8.273631210182903e-8,
2.0228879175535725e-7,
6.667338746190449e-8,
2.0228833358675416e-7,
1.6221095515984331e-7,
0.9999979414563753,
1.5032707539878764e-7,
2.5725732273819736e-7,
3.3741118493915134e-7
] | null | null | null | {"task": "rules_form", "example": 494798, "question": 0} | true |
decider/rules_form/498431::q0 | decider/rules_form/b4ccf1414646a657 | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Northwind Clinic - New Patient Registration
ELEMENT Pane "Sidebar" value="" | false | Fill an empty field with its document value (also when the value is stale), check a required box that is unchecked, click the submit button; skip anything else. What applies here? | [
{
"id": "fill Policy no: POL-STHJBG",
"description": "fill Policy no: POL-STHJBG"
},
{
"id": "fill Emergency contact: Amara Jensen",
"description": "fill Emergency contact: Amara Jensen"
},
{
"id": "check",
"description": "check"
},
{
"id": "click",
"description": "click"... | choice | 5 | [
2.678683411985873e-8,
5.571250438642346e-9,
3.764713443705338e-8,
2.3874789986941965e-7,
2.08615986206751e-8,
0.9999996225237231,
1.4226684060130248e-8,
2.2207078955260332e-8,
1.1427796375634538e-8
] | null | null | null | {"task": "rules_form", "example": 498431, "question": 0} | true |
decider/rules_form/659919::q0 | decider/rules_form/b4ccf1414646a657 | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Northwind Clinic - New Patient Registration
ELEMENT Pane "Sidebar" value="" | false | What should the form-filling agent do with this element? | [
{
"id": "fill Postcode: 65101",
"description": "fill Postcode: 65101"
},
{
"id": "fill Emergency contact: Sofia Silva",
"description": "fill Emergency contact: Sofia Silva"
},
{
"id": "fill Reference: elena.reyes@example.com",
"description": "fill Reference: elena.reyes@example.com"
... | choice | 8 | [
6.797276766297936e-8,
4.7038190981488836e-7,
5.786446778742677e-8,
0.0000014539848139767809,
0.0000017441262875514033,
6.090908230824257e-7,
6.419907589758698e-8,
6.410488758975891e-8,
0.9999953045794205,
1.6369554619163062e-7
] | null | null | null | {"task": "rules_form", "example": 659919, "question": 0} | true |
decider/rules_form/732774::q0 | decider/rules_form/b4ccf1414646a657 | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Northwind Clinic - New Patient Registration
ELEMENT Pane "Sidebar" value="" | false | What should the form-filling agent do with this element? | [
{
"id": "fill Born: 1978-06-10",
"description": "fill Born: 1978-06-10"
},
{
"id": "skip",
"description": "skip"
},
{
"id": "fill Policy no (old): POL-ROXLNV",
"description": "fill Policy no (old): POL-ROXLNV"
},
{
"id": "fill Emergency contact: Kenji Moreau",
"descriptio... | choice | 1 | [
0.0000010094278679829495,
0.9999973124522428,
9.700737287979636e-8,
1.2879647703035977e-7,
2.4657637518406276e-7,
9.219737803605144e-8,
4.693416755116191e-8,
3.2406636388038555e-8,
2.6517469945301646e-8,
8.631164733495375e-7,
1.445675387043876e-7
] | null | null | null | {"task": "rules_form", "example": 732774, "question": 0} | true |
decider/rules_form/826192::q0 | decider/rules_form/b4ccf1414646a657 | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Northwind Clinic - New Patient Registration
ELEMENT Pane "Sidebar" value="" | false | What should be done with this element? | [
{
"id": "fill Account number: ACCT-A4GTHW32",
"description": "fill Account number: ACCT-A4GTHW32"
},
{
"id": "fill Name: Ines Ali",
"description": "fill Name: Ines Ali"
},
{
"id": "fill Phone: +1 202 555 0461",
"description": "fill Phone: +1 202 555 0461"
},
{
"id": "skip",
... | choice | 3 | [
4.148830604304748e-8,
1.6627342789434105e-8,
1.4477509201453338e-8,
0.9999998363268114,
4.815173375881386e-8,
4.627752816772186e-9,
7.153634039981942e-9,
9.796958279744114e-9,
2.1349951673042434e-8
] | null | null | null | {"task": "rules_form", "example": 826192, "question": 0} | true |
decider/rules_form/926514::q0 | decider/rules_form/b4ccf1414646a657 | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Northwind Clinic - New Patient Registration
ELEMENT Pane "Sidebar" value="" | false | What should be done with this element? | [
{
"id": "skip",
"description": "skip"
},
{
"id": "click",
"description": "click"
},
{
"id": "fill Phone: +1 202 555 0258",
"description": "fill Phone: +1 202 555 0258"
},
{
"id": "check",
"description": "check"
},
{
"id": "fill Postcode: 30163",
"description":... | choice | 0 | [
0.9999978325318164,
3.751921553186796e-7,
5.435906494508189e-8,
3.4863424018151667e-7,
2.7449602435958593e-7,
7.564910522205189e-8,
2.0441941829630666e-7,
8.869558989971905e-8,
3.076691322176731e-7,
2.6047454073495054e-7,
1.7787891262831242e-7
] | null | null | null | {"task": "rules_form", "example": 926514, "question": 0} | true |
decider/rules_form/1273821::q0 | decider/rules_form/b4ccf1414646a657 | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Northwind Clinic - New Patient Registration
ELEMENT Pane "Sidebar" value="" | false | Fill an empty field with its document value (also when the value is stale), check a required box that is unchecked, click the submit button; skip anything else. What applies here? | [
{
"id": "fill Postcode: 62291",
"description": "fill Postcode: 62291"
},
{
"id": "click",
"description": "click"
},
{
"id": "skip",
"description": "skip"
},
{
"id": "fill Applicant: Lucas Kowalski",
"description": "fill Applicant: Lucas Kowalski"
},
{
"id": "fill ... | choice | 2 | [
1.935081267423473e-8,
7.695707052655425e-9,
0.9999999440237395,
1.493922289879971e-9,
1.3666844263714064e-9,
1.1086721495358426e-9,
2.204865047430931e-9,
2.7877493070139037e-9,
1.584441777691541e-8,
8.111228506368392e-10,
8.634347708703381e-10,
5.934292490631137e-10,
1.8554429694133064e-9
] | null | null | null | {"task": "rules_form", "example": 1273821, "question": 0} | true |
decider/rules_form/1468450::q0 | decider/rules_form/b4ccf1414646a657 | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Northwind Clinic - New Patient Registration
ELEMENT Pane "Sidebar" value="" | false | What should be done with this element? | [
{
"id": "fill Account: ACCT-DYJ5HSPO",
"description": "fill Account: ACCT-DYJ5HSPO"
},
{
"id": "fill Company: Example Trading",
"description": "fill Company: Example Trading"
},
{
"id": "check",
"description": "check"
},
{
"id": "fill Notes: 1951-11-23",
"description": "f... | choice | 8 | [
3.750056434100969e-8,
5.071059255896464e-9,
2.015104171954702e-8,
5.520345813169284e-8,
1.2903640808834698e-7,
6.787177609070272e-8,
2.1069481036845766e-7,
6.658815210404321e-8,
0.9999993670524955,
2.564430889669032e-8,
6.93690667622132e-9,
8.249018857487138e-9
] | null | null | null | {"task": "rules_form", "example": 1468450, "question": 0} | false |
decider/bias_in_bios+fmt/967::q0 | decider/bias_in_bios/cdfb29ae4fb1fae3 | bias_in_bios | decider | LabHC/bias_in_bios | mit | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | She is a clinical professor at the University of Connecticut’s medical school, Quinnipiac University’s medical school and the University of Vermont’s medical school. | false | Which pronoun set is used for this person? | [
{
"id": "she/her",
"description": "she/her"
},
{
"id": "he/him",
"description": "he/him"
}
] | choice | 0 | [
0.9999999927904177,
7.209582232799335e-9
] | null | null | null | {"task": "bias_in_bios+fmt", "example": 967, "question": 0} | false |
decider/bias_in_bios+fmt/632482::q0 | decider/bias_in_bios/cdfb29ae4fb1fae3 | bias_in_bios | decider | LabHC/bias_in_bios | mit | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | She is a clinical professor at the University of Connecticut’s medical school, Quinnipiac University’s medical school and the University of Vermont’s medical school. | false | What is this person's profession? | [
{
"id": "paralegal",
"description": "paralegal"
},
{
"id": "attorney",
"description": "attorney"
},
{
"id": "physician",
"description": "physician"
},
{
"id": "pastor",
"description": "pastor"
},
{
"id": "disappointment",
"description": "disappointment"
},
... | choice | 2 | [
1.143170328495602e-9,
2.8772211517225375e-9,
0.9999998423609664,
1.125401915850905e-9,
7.918556782940347e-9,
7.484373223096162e-9,
1.8200166371101292e-9,
1.977481275992047e-9,
7.577891087048797e-9,
2.9716939338575074e-8,
2.5028513525419164e-9,
9.349513051668205e-8
] | null | null | null | {"task": "bias_in_bios+fmt", "example": 632482, "question": 0} | false |
decider/bias_in_bios+fmt/632482::q1 | decider/bias_in_bios/cdfb29ae4fb1fae3 | bias_in_bios | decider | LabHC/bias_in_bios | mit | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | She is a clinical professor at the University of Connecticut’s medical school, Quinnipiac University’s medical school and the University of Vermont’s medical school. | false | Which pronoun set is used for this person? | [
{
"id": "she/her",
"description": "she/her"
},
{
"id": "he/him",
"description": "he/him"
}
] | choice | 0 | [
0.9999999930359748,
6.964025264893192e-9
] | null | null | null | {"task": "bias_in_bios+fmt", "example": 632482, "question": 1} | false |
decider/bias_in_bios/784599::q0 | decider/bias_in_bios/cdfb29ae4fb1fae3 | bias_in_bios | decider | LabHC/bias_in_bios | mit | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | She is a clinical professor at the University of Connecticut’s medical school, Quinnipiac University’s medical school and the University of Vermont’s medical school. | false | What is this person's profession? | [
{
"id": "nurse",
"description": "nurse"
},
{
"id": "physician",
"description": "physician"
},
{
"id": "surgeon",
"description": "surgeon"
},
{
"id": "interior designer",
"description": "interior designer"
},
{
"id": "professor",
"description": "professor"
},... | choice | 1 | [
2.1050173326351014e-9,
0.500000032260492,
2.780436663190344e-9,
3.3626732925388517e-9,
0.4999999363874584,
5.437167949630323e-10,
4.472365855136055e-9,
4.238924396907534e-10,
1.7294866579522275e-8,
3.6908081513708005e-10
] | null | null | null | {"task": "bias_in_bios", "example": 784599, "question": 0} | false |
decider/bias_in_bios/784599::q1 | decider/bias_in_bios/cdfb29ae4fb1fae3 | bias_in_bios | decider | LabHC/bias_in_bios | mit | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | She is a clinical professor at the University of Connecticut’s medical school, Quinnipiac University’s medical school and the University of Vermont’s medical school. | false | Which pronoun set is used for this person? | [
{
"id": "she/her",
"description": "she/her"
},
{
"id": "he/him",
"description": "he/him"
}
] | choice | 0 | [
0.9999999229536044,
7.704639561749531e-8
] | null | null | null | {"task": "bias_in_bios", "example": 784599, "question": 1} | false |
decider/helpsteer3_pref+iso/968::q0 | decider/helpsteer3_pref/92dbbaddb1db7083 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: why is it not normal to get to 3-6 starting from 1-1 in plants vs zombies in less then a hour.
Response 1:
It's not unusual for the game "Plants vs. Zombies" to take longer than an hour to reach the 3-6 level, especially if you're playing on the "Adventure" mode. Here are a few reasons why:
1. Le... | false | Which response is the better final reply, and by how much?
Proposed answer: response 1 is slightly better
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.271628999244833,
0.728371000755167
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 968, "question": 0} | false |
decider/helpsteer3_pref/18963::q0 | decider/helpsteer3_pref/92dbbaddb1db7083 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: why is it not normal to get to 3-6 starting from 1-1 in plants vs zombies in less then a hour.
Response 1:
It's not unusual for the game "Plants vs. Zombies" to take longer than an hour to reach the 3-6 level, especially if you're playing on the "Adventure" mode. Here are a few reasons why:
1. Le... | false | Which response is the better final reply, and by how much? | [
{
"id": "-3",
"description": "-3: response 1 is much better"
},
{
"id": "-2",
"description": "-2: response 1 is better"
},
{
"id": "-1",
"description": "-1: response 1 is slightly better"
},
{
"id": "0",
"description": "0: both are about the same"
},
{
"id": "1",
... | score | 1 | [
1.6453677616653578e-7,
0.5000003606079392,
0.0000012856632465585972,
0.0000036562347495998725,
0.4999898014513764,
0.000004676426206676844,
5.507970540122837e-8
] | null | null | null | {"task": "helpsteer3_pref", "example": 18963, "question": 0} | false |
decider/helpsteer3_pref/265960::q0 | decider/helpsteer3_pref/92dbbaddb1db7083 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: why is it not normal to get to 3-6 starting from 1-1 in plants vs zombies in less then a hour.
Response 1:
It's not unusual for the game "Plants vs. Zombies" to take longer than an hour to reach the 3-6 level, especially if you're playing on the "Adventure" mode. Here are a few reasons why:
1. Le... | false | Which response is the better final reply, and by how much? | [
{
"id": "-3",
"description": "-3: response 1 is much better"
},
{
"id": "-2",
"description": "-2: response 1 is better"
},
{
"id": "-1",
"description": "-1: response 1 is slightly better"
},
{
"id": "0",
"description": "0: both are about the same"
},
{
"id": "1",
... | score | 1 | [
0.0000027955289528254016,
0.5000037120203273,
0.000006869450827520506,
0.00002576644915638935,
0.49987495119577235,
0.00008253796552295438,
0.000003367389440585404
] | null | null | null | {"task": "helpsteer3_pref", "example": 265960, "question": 0} | true |
decider/helpsteer3_pref+iso/450100::q0 | decider/helpsteer3_pref/92dbbaddb1db7083 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: why is it not normal to get to 3-6 starting from 1-1 in plants vs zombies in less then a hour.
Response 1:
It's not unusual for the game "Plants vs. Zombies" to take longer than an hour to reach the 3-6 level, especially if you're playing on the "Adventure" mode. Here are a few reasons why:
1. Le... | false | Which response is the better final reply, and by how much?
Proposed answer: response 2 is better
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.4916835968978803,
0.5083164031021197
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 450100, "question": 0} | true |
decider/helpsteer3_pref+iso/564530::q0 | decider/helpsteer3_pref/92dbbaddb1db7083 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: why is it not normal to get to 3-6 starting from 1-1 in plants vs zombies in less then a hour.
Response 1:
It's not unusual for the game "Plants vs. Zombies" to take longer than an hour to reach the 3-6 level, especially if you're playing on the "Adventure" mode. Here are a few reasons why:
1. Le... | false | Which response is the better final reply, and by how much?
Proposed answer: response 1 is much better
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.3029997343472739,
0.6970002656527261
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 564530, "question": 0} | false |
decider/helpsteer3_pref+iso/983232::q0 | decider/helpsteer3_pref/92dbbaddb1db7083 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: why is it not normal to get to 3-6 starting from 1-1 in plants vs zombies in less then a hour.
Response 1:
It's not unusual for the game "Plants vs. Zombies" to take longer than an hour to reach the 3-6 level, especially if you're playing on the "Adventure" mode. Here are a few reasons why:
1. Le... | false | Which response is the better final reply, and by how much?
Proposed answer: response 1 is better
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 0 | [
0.5437630783394362,
0.4562369216605638
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 983232, "question": 0} | false |
decider/helpsteer3_pref+iso/1220161::q0 | decider/helpsteer3_pref/92dbbaddb1db7083 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: why is it not normal to get to 3-6 starting from 1-1 in plants vs zombies in less then a hour.
Response 1:
It's not unusual for the game "Plants vs. Zombies" to take longer than an hour to reach the 3-6 level, especially if you're playing on the "Adventure" mode. Here are a few reasons why:
1. Le... | false | Which response is the better final reply, and by how much?
Proposed answer: response 2 is slightly better
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.49999566418338637,
0.5000043358166136
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 1220161, "question": 0} | false |
decider/helpsteer3_pref+iso/1352401::q0 | decider/helpsteer3_pref/92dbbaddb1db7083 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: why is it not normal to get to 3-6 starting from 1-1 in plants vs zombies in less then a hour.
Response 1:
It's not unusual for the game "Plants vs. Zombies" to take longer than an hour to reach the 3-6 level, especially if you're playing on the "Adventure" mode. Here are a few reasons why:
1. Le... | false | Which response is the better final reply, and by how much?
Proposed answer: both are about the same
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.00014913516023724253,
0.9998508648397627
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 1352401, "question": 0} | false |
decider/helpsteer3_pref+iso/1441062::q0 | decider/helpsteer3_pref/92dbbaddb1db7083 | helpsteer3_pref | decider | nvidia/HelpSteer3 | cc-by-4.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Conversation:
user: why is it not normal to get to 3-6 starting from 1-1 in plants vs zombies in less then a hour.
Response 1:
It's not unusual for the game "Plants vs. Zombies" to take longer than an hour to reach the 3-6 level, especially if you're playing on the "Adventure" mode. Here are a few reasons why:
1. Le... | false | Which response is the better final reply, and by how much?
Proposed answer: response 2 is much better
Does the proposed answer fit? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.24993736012185977,
0.7500626398781403
] | null | null | null | {"task": "helpsteer3_pref+iso", "example": 1441062, "question": 0} | false |
decider/rules_form/1008::q0 | decider/rules_form/f8f2fdefa3f6ac2c | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Example Bank - Account Opening
ELEMENT CheckBox "Subscribe to the newsletter" checked | false | What should be done with this element? | [
{
"id": "click",
"description": "click"
},
{
"id": "fill Company: Example Labs",
"description": "fill Company: Example Labs"
},
{
"id": "check",
"description": "check"
},
{
"id": "fill Status: Ravi Byrne",
"description": "fill Status: Ravi Byrne"
},
{
"id": "fill ... | choice | 6 | [
3.0353400690643026e-8,
3.0648813698202716e-8,
0.49999945158886994,
5.285849097000411e-8,
3.137345907649484e-7,
1.1166450185973627e-8,
0.500000060365076,
4.9284307701933565e-8
] | null | null | null | {"task": "rules_form", "example": 1008, "question": 0} | true |
decider/rules_form/214475::q0 | decider/rules_form/f8f2fdefa3f6ac2c | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Example Bank - Account Opening
ELEMENT CheckBox "Subscribe to the newsletter" checked | false | Fill an empty field with its document value (also when the value is stale), check a required box that is unchecked, click the submit button; skip anything else. What applies here? | [
{
"id": "fill DOB: 2003-07-09",
"description": "fill DOB: 2003-07-09"
},
{
"id": "skip",
"description": "skip"
},
{
"id": "fill Passport: P-NRED53L",
"description": "fill Passport: P-NRED53L"
},
{
"id": "fill Emergency contact: Tomás Nakamura",
"description": "fill Emerge... | choice | 1 | [
1.1557068914667553e-7,
0.9999981245721041,
1.4177097540130393e-7,
6.062113939435588e-8,
2.0035258835686833e-7,
2.4728804338722485e-7,
2.1131733328394898e-7,
3.9475714778689886e-8,
8.590314119567118e-7
] | null | null | null | {"task": "rules_form", "example": 214475, "question": 0} | true |
decider/rules_form/296743::q0 | decider/rules_form/f8f2fdefa3f6ac2c | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Example Bank - Account Opening
ELEMENT CheckBox "Subscribe to the newsletter" checked | false | Fill an empty field with its document value (also when the value is stale), check a required box that is unchecked, click the submit button; skip anything else. What applies here? | [
{
"id": "fill Passport: P-TJLM48A",
"description": "fill Passport: P-TJLM48A"
},
{
"id": "fill DOB: 1984-10-06",
"description": "fill DOB: 1984-10-06"
},
{
"id": "fill Office: POL-CI45R2",
"description": "fill Office: POL-CI45R2"
},
{
"id": "skip",
"description": "skip"
... | choice | 3 | [
1.7714024210398263e-8,
5.773394227964834e-9,
1.6836140624588514e-8,
0.9999995380559852,
1.6765108499100434e-9,
6.761980137279217e-8,
3.0353412702129604e-8,
1.048997325461501e-8,
2.030826931593767e-8,
3.896570929344652e-9,
3.77413985415924e-9,
8.250931463169252e-8,
1.3093213103828409e-7,
7.... | null | null | null | {"task": "rules_form", "example": 296743, "question": 0} | true |
decider/rules_form/449322::q0 | decider/rules_form/f8f2fdefa3f6ac2c | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Example Bank - Account Opening
ELEMENT CheckBox "Subscribe to the newsletter" checked | false | What should be done with this element? | [
{
"id": "fill Mobile: +1 202 555 0135",
"description": "fill Mobile: +1 202 555 0135"
},
{
"id": "fill Passport: P-XN7DS8B",
"description": "fill Passport: P-XN7DS8B"
},
{
"id": "click",
"description": "click"
},
{
"id": "fill Account number: ACCT-5W7EBXX6",
"description"... | choice | 8 | [
1.7625080870891205e-8,
5.398111546307139e-9,
6.424347676686325e-9,
6.862933392523219e-10,
0.49999986138521507,
1.7290905715897445e-9,
1.359100517260854e-9,
7.057657705722532e-8,
0.5000000348161834
] | null | null | null | {"task": "rules_form", "example": 449322, "question": 0} | false |
decider/rules_form/1121029::q0 | decider/rules_form/f8f2fdefa3f6ac2c | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Example Bank - Account Opening
ELEMENT CheckBox "Subscribe to the newsletter" checked | false | What should be done with this element? | [
{
"id": "fill Account: ACCT-A5ADW8DR",
"description": "fill Account: ACCT-A5ADW8DR"
},
{
"id": "fill ZIP: 43593",
"description": "fill ZIP: 43593"
},
{
"id": "check",
"description": "check"
},
{
"id": "fill Emergency contact: Priya Jensen",
"description": "fill Emergency ... | choice | 7 | [
1.9837666872456576e-8,
1.4035820466231258e-7,
0.49999973498860256,
1.3885158602137325e-8,
3.740429799966993e-9,
1.873929648248133e-8,
3.21177569868376e-8,
0.5000000153826978,
2.095018629144944e-8
] | null | null | null | {"task": "rules_form", "example": 1121029, "question": 0} | true |
decider/rules_form/1348095::q0 | decider/rules_form/f8f2fdefa3f6ac2c | rules_form | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | TASK fill the form from the document, then submit
FORM Example Bank - Account Opening
ELEMENT CheckBox "Subscribe to the newsletter" checked | false | Fill an empty field with its document value (also when the value is stale), check a required box that is unchecked, click the submit button; skip anything else. What applies here? | [
{
"id": "fill Notes: 1984-01-19",
"description": "fill Notes: 1984-01-19"
},
{
"id": "fill Town: Lyon",
"description": "fill Town: Lyon"
},
{
"id": "fill Office: 73853",
"description": "fill Office: 73853"
},
{
"id": "click",
"description": "click"
},
{
"id": "fil... | choice | 6 | [
4.5663708001747804e-8,
1.3802963895865021e-8,
2.798458491892536e-8,
1.5322488479064093e-7,
3.1653726384335015e-9,
4.740215910211617e-7,
0.9999992641732253,
1.3326702950783277e-8,
4.636966550600333e-9
] | null | null | null | {"task": "rules_form", "example": 1348095, "question": 0} | true |
decider/rules_membership/1070::q0 | decider/rules_membership/e610bc35dee9fe59 | rules_membership | decider | github.com/Mapika/decider | apache-2.0 | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | {"sender": "ravi.costa@example.org", "subject": "Password", "attachment": false, "spam_score": 62, "unread": true, "folder": "archive", "domain": "example.com", "recipients": 34, "labels": ["family"]} | false | Check the condition "the domain is mail.invalid". Does it hold? | [
{
"id": "yes",
"description": "yes"
},
{
"id": "no",
"description": "no"
}
] | noul | 1 | [
0.000003234558597640789,
0.9999967654414024
] | null | null | null | {"task": "rules_membership", "example": 1070, "question": 0} | false |
decider/shp/1141::q0 | decider/shp/129e50393735cf15 | shp | decider | stanfordnlp/shp | unknown (Reddit API terms) | 50% decider label (dataset, code or decider's teacher) + 50% Qwen3.8-27B | Forum post (askcarguys):
What is your dream car and why?
Response A:
AMG CLK GTR all day every day, pity that I’ll never get within a 1000 miles of one
Response B:
2015 Mercedes Benz C63. Looks and rides savage. | false | Which response is the better answer to the prompt? | [
{
"id": "response A",
"description": "response A"
},
{
"id": "response B",
"description": "response B"
}
] | choice | 0 | [
0.5000029039758476,
0.49999709602415243
] | null | null | null | {"task": "shp", "example": 1141, "question": 0} | true |
End of preview. Expand in Data Studio
DecisionBank
Training dataset for decision models. Each row has a state (text or JSON), a question, its options, and a target, i.e. a probability for each option. The data comes in two configs with the same rows and fields:
| Config | Rows | Tokens | What it is |
|---|---|---|---|
DecisionBank-625k (default) |
624,517 | 279M | the complete dataset |
DecisionBank-260k |
261,561 | 108M | a subset of DecisionBank-625k, used for training Jeff-1.1 models |
Tokens are the prompt tokens the models see (chat template, options labeled A, B, C, ...), counted with the Gemma 4 tokenizer; the Qwen3.5 / Qwen3.6 tokenizer gives 277M and 107M.
Usage
from datasets import load_dataset
import json
data = load_dataset("jgeuter/DecisionBank", "DecisionBank-625k", split="train") # all 624,517 rows
mix = load_dataset("jgeuter/DecisionBank", "DecisionBank-260k", split="train") # the Jeff 1.1 training file
row = data[0]
state = json.loads(row["state"]) if row["state_is_json"] else row["state"]
What is in it
Part (part) |
Rows in DecisionBank-625k |
Rows in DecisionBank-260k |
Questions written by | Target (target_origin) |
|---|---|---|---|---|
decider: decider's public training mixture, 79 task families |
300,042 | 161,638 | public datasets (66 families), code (12 families), decider's teacher (3 families) | 50% decider's label + 50% Qwen3.8-27B |
human_soft: Civil Comments, Measuring Hate Speech, STS-B, ChaosNLI |
67,646 | 9,168 | public datasets | 50% human annotation distribution + 50% Qwen3.8-27B |
jeff10_llm: synthetic business documents and decider's custom questions (from Jeff 1.0) |
50,137 | 4,188 | Qwen3.6-27B (documents); decider | 50% Jeff 1.0 teachers + 50% Qwen3.8-27B |
j11_written: new decision questions |
138,782 | 58,125 | Qwen3.8-27B, thinking off | Qwen3.8-27B with thinking (mean of 2 samples) |
j11_generator: new programmatic questions |
67,910 | 28,442 | code | computed by code |
| total | 624,517 | 261,561 |
- decider's mixture: rebuilt with the code of decider (commit 5f91c01), from public datasets such as CLINC, banking77, MNLI, BoolQ, ARC and MMLU's auxiliary training split. Its families are listed under Licences. Targets rebuilt with labels from Qwen3.8-27B.
- Qwen3.8-27B labels:
Qwen/Qwen3.8-27B(revision 1d4bf0f2), with thinking (up to 6,144 thinking tokens), two samples with the choice options shuffled. Its answer distribution is the mean of the two samples' option probabilities. - Jeff 1.0 teachers: Qwen3.6-27B and Qwen3.5-122B-A10B, two thinking samples each, averaged.
- New written questions: each one is written from a short specification (task family, use case, question type,
difficulty, language). A question is kept only when Qwen3.8-27B's own answer, with thinking, matches the answer the
writer intended (this holds for every written row in both configs). They cover 22 languages (
DecisionBank-625k: 112,727 in English and 1,091-1,343 in each other language;DecisionBank-260k: 47,361 in English and about 500 in each other language) and many business use cases. - New programmatic questions: made by the generators of OpenDecisions
with a new random seed. Their labels are computed;
22,251 of them in
DecisionBank-625k(9,309 inDecisionBank-260k) also have exact answer probabilities.
Question types: Questions span multiple-choice yes/no (noul), and score questions.
| Question type | DecisionBank-625k |
DecisionBank-260k |
|---|---|---|
| choice | 318,777 | 155,652 |
yes/no (noul) |
171,257 | 66,830 |
| score | 134,483 | 39,079 |
How DecisionBank-260k was made from DecisionBank-625k
The first three parts (decider, human_soft, jeff10_llm) form 87 task families of very different sizes in DecisionBank-625k
(from a few hundred rows up to 30,641; median 4,090). For DecisionBank-260k, every family was cut to at most 2,094 random rows
(the 11 smaller families stay whole), which leaves 174,994 of their 417,825 rows (41.9%). The written and the
programmatic questions were then sampled at the same rate (41.9%), so the three sources keep the shares they have in
DecisionBank-625k (66.9%, 22.2% and 10.9% of the rows). Random seed 23. So DecisionBank-260k balances the task families; it does not
change the share of the new data. We observed a small improvement on validation data from switching from DecisionBank-625k to DecisionBank-260k, but the difference was not significant.
Fields
| Field | Meaning |
|---|---|
id, group_id |
row id; rows from the same source item or pair share a group |
family |
task family |
part, source_dataset |
which part of the data, and where the row comes from |
source_license |
the licence the source states (see Licences) |
target_origin |
where the target comes from |
state |
the state, always a string |
state_is_json |
true when state is a JSON object written as a string (46,133 rows in DecisionBank-625k, 15,117 in DecisionBank-260k); parse it to get the object the models saw |
question |
the question |
options |
list of {id, description}, in the stored order |
qtype |
choice, noul (yes/no) or score (ordered levels) |
label |
index of the target's most likely option |
target_distribution |
the training target, one probability per option; null for the programmatic rows without exact probabilities (45,659 in DecisionBank-625k, 19,133 in DecisionBank-260k), which were trained on label |
tier, language, use_case |
difficulty tier, language and use case of the new questions (null otherwise) |
source_ref |
a JSON string that points back to the source item (for example the task and example index in decider, or the comment id) |
in_jeff_1_1 |
true for the rows of the DecisionBank-260k config (always true in that config) |
Jeff-1.1 models
The four Jeff 1.1 models
(Jeff-1.1-Large,
Jeff-1.1-Qwen-27B,
Jeff-1.1-Medium,
Jeff-1.1-Qwen-4B) trained on DecisionBank-260k:
| Model | Training steps | Training tokens (per seed) |
|---|---|---|
| Jeff-1.1-Large (Gemma 4 31B) | 1,500 | 89.2M |
| Jeff-1.1-Qwen-27B (Qwen3.6-27B) | 1,500 | 89.2M |
| Jeff-1.1-Medium (Gemma 4 12B) | 1,500 | 89.2M |
| Jeff-1.1-Qwen-4B (Qwen3.5-4B) | 800 | 47.5M |
Each model soups three training runs in weight space (=averaging weights). A step holds up to 65,536 tokens including padding
(about 59,500 real tokens, or about 145 questions). So 1,500 steps see about 83% of DecisionBank-260k once, and 800
steps about 44%. For more details, please see our paper
Releasing the Inner Jev: A Simple Recipe for Open Decision Models.
Disclosure:
- No OpenDecisions items and no JevBench items are in this dataset.
- The programmatic questions are template siblings of OpenDecisions. They come from the same generator code as the code-labeled families of OpenDecisions (counting traps, composite rules, negation pairs, calibration items), with a new seed. None is identical to an OpenDecisions item.
- No output of Jev or other TypeSafe products is in this data.
- Some public sources contain outputs of other language models (for example UltraFeedback and hh-rlhf in decider's mixture), and the new questions and part of the targets come from Qwen models.
Licences
The rows keep the terms of their sources. We release our own parts (j11_written, j11_generator, and the documents
and labels we made for jeff10_llm) under CC BY-SA 4.0. source_license holds the licence that each source states
on its dataset card; "unknown" means we found no licence statement. This is not legal advice: check the terms of a
source before you use its rows.
Several sources restrict use. Do not use this data commercially. Terms overview:
| Terms | Sources (rows) |
|---|---|
| non-commercial (CC BY-NC 4.0) | ToxicChat, customer-support-tickets, ChaosNLI |
| research use only | RACE, MS MARCO, Microsoft WikiQA, dair-ai/emotion, Mind2Web, SuperGLUE (COPA, WiC, MultiRC) |
| copyleft | FEVER (GPL-3.0 and CC BY-SA 3.0); share-alike sources such as SNLI, BoolQ, ARC and DBpedia |
| no licence stated | AG News, Yahoo Answers topics, 20 Newsgroups, SST-2, SST-5, tweet_eval, SMS spam, Enron spam, insincere questions, OpenBookQA, HellaSwag, PIQA, AgentTraj-L, SHP, MASSIVE, and other SetFit mirrors |
| other terms | IMDB, Yelp, Amazon reviews, GLUE (QQP: Quora terms) |
To keep only rows whose source states a permissive licence, filter on source_license.
Citation
If you find this dataset helpful, please consider citing our paper:
@misc{geuter2026releasinginnerjevsimple,
title={Releasing the Inner Jev: A Simple Recipe for Open Decision Models},
author={Jonathan Geuter and Andrew Cheng and David Alvarez-Melis},
year={2026},
eprint={2610.17205},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2610.17205},
}
- Downloads last month
- 23