Datasets:

Alberto1231
/

thoughttrace_modified

prompt string	target string	date string	conversation_id string	user_id string	source_model string	model_provider string	human_message string	assistant_message string	source_turn_index int64	target_message_id string	target_role string	source_target_type string	n_context_turns int64	prompt_token_count int64	target_token_count int64	target_word_count int64	gender string	age_group string	education string	occupation string	lm_frequency_use string	ai_purposes string	cohort_gender bool	cohort_age_group bool	cohort_education bool	cohort_lm_frequency_use bool	cohort_source_model bool
Conversation: Human: Hi Assistant: Hi! How can I help? Human:	I need help	2026-04-06T20:02:56.855000Z	user266_task2_conversation1	user266	GPT-5.4	OpenAI	Hi	Hi! How can I help?	2	1775505776855	human	user	2	18	3	3	Female	18-24	Undergraduate	Manager	Many times a day	School and work	false	false	false	false	false
Conversation: Human: Good morning Assistant: Good morning! ☀️ How's your day starting off? Human:	Good good, how about you?	2026-04-06T04:20:08.458000Z	user170_task2_conversation1	user170	Grok 4.20	xAI	Good morning	Good morning! ☀️ How's your day starting off?	2	1775449208458	human	user	2	25	7	5	Male	25-34	Graduate	Lawyer	A couple of times	Personal research, grammar.	false	false	false	false	false
Conversation: Human: Hello Assistant: Hello! How can I help you today? Human:	Give me suggestions for interesting places to visit in Europe.	2026-03-18T23:11:11.312000Z	user65_task1_conversation1	user65	Gemini 3.1 Pro Preview	Google	Hello	Hello! How can I help you today?	2	1773875471312	human	user	2	20	11	10	Male	35-44	High school	quality technician	Once a week	To help with personal research, so that it can be faster.	false	false	false	false	false
Conversation: Human: Hello Assistant: Hello! How can I help you today? Human:	I would like to have a barbecue today.	2026-04-12T05:16:49.599000Z	user861_task1_conversation1	user861	Qwen3.6 Plus	Qwen	Hello	Hello! How can I help you today?	2	1775971009599	human	user	2	20	9	8	Male	45-54	Graduate	Civil Engineer	Once a day	Working and learning	false	false	false	false	false
Conversation: Human: hi Assistant: Hello! How can I help you today? Human:	i want to plan a trip	2026-04-12T17:17:36.704000Z	user920_task1_conversation1	user920	Gemma 4 26B A4B	Google	hi	Hello! How can I help you today?	2	1776014256704	human	user	2	20	6	6	Male	25-34	Other	Data scientist	Many times a day	working and learning new things	false	false	false	false	false

ThoughtTrace Modified

A next-human-turn benchmark derived from SCAI-JHU/ThoughtTrace at pinned revision 0420f3d8499e477098aac7771fe9c066f2340fb3.

Non-negotiable target contract

Every scored target is copied from a source message whose type is exactly user, immediately following a source message whose type is exactly assistant. The assistant message is conditioning context, never the target:

Human: previous human message
Assistant: source LLM response
Human: <genuine human target>

The builder validates this transition against the pinned source for every row. Private reasons, reactions, post-task summaries, task expectations, and demographics are never included in the model prompt.

Splits and selection

train: 5 demonstrations from distinct participants
test: 965 scored targets from distinct participants
One row per participant across train and test
Every normalized target string is globally unique
Human-message date range: 2026-03-18T15:13:24.848000Z through 2026-04-14T12:26:00.404000Z
Both the source conversation and all three messages in the selected user-assistant-user transition must be on or after 2023-01-01T00:00:00Z
All 20 source assistant models remain eligible
Five demonstrations are selected for compactness and diversity across source model, gender, age, education, usage frequency, and surface response form
The full test split is intentionally a near-census, not an artificially equalized demographic sample
Selection never oversamples or duplicates people from small groups

The model receives only the immediately preceding human message and assistant response. Complete five-shot prompts fit a 2,048-token model context after reserving 256 tokens for generation with the pinned MARIN tokenizer.

Metadata and subgroup analysis

Self-reported gender, normalized age group, education, occupation, AI-use frequency, and AI purposes are retained as analysis metadata. They never enter the prompt. Invalid ages and unanswered fields become Missing; identities are never inferred from conversation text.

The natural full split is allowed to be unbalanced. For convenient matched-N ablations, immutable cohort_<attribute> flags identify deterministic equal-N subsets for sufficiently supported levels:

gender: 393 respondents per included level (Female, Male)
age_group: 43 respondents per included level (18-24, 25-34, 35-44, 45-54)
education: 97 respondents per included level (Graduate, High school, Undergraduate)
lm_frequency_use: 49 respondents per included level (A couple of times, Many times a day, Once a day, Once a week)
source_model: 30 respondents per included level (Claude Opus 4.6, Claude Sonnet 4.6, GPT-4o-mini, GPT-5.4, Gemini 3 Flash Preview, Gemini 3.1 Pro Preview, Gemma 4 26B A4B, Grok 4.1 Fast, Grok 4.20, Kimi K2.5, Llama 3.3 70B Instruct, MiMo-V2-Pro, MiniMax M2.7)

These are descriptive cohorts, not causal controls. Attributes, task content, and assigned source assistant model may remain correlated.

Downloads last month: 103