Dataset Viewer
Auto-converted to Parquet Duplicate
trace_id
stringclasses
4 values
trace_type
stringclasses
1 value
created_date
timestamp[s]date
2026-06-08 00:00:00
2026-06-08 00:00:00
source
stringclasses
1 value
space_url
stringclasses
1 value
runtime_mode
stringclasses
1 value
runtime_note
stringclasses
1 value
input
dict
output
dict
validation
dict
privacy
dict
exam-panic-live-01
live_model_run
2026-06-08T00:00:00
Exam Panic Rescue Gradio Space (live generate endpoint)
https://huggingface.co/spaces/build-small-hackathon/exam-panic-rescue
zerogpu_minicpm_live
Generated with openbmb/MiniCPM4.1-8B (8B) on CUDA/ZeroGPU. Drill source: MiniCPM-generated drills.
{ "scenario_name": "biology panic", "student_name": "Mira", "subject": "Biology: cell division", "time_left_minutes": 45, "exam_format": "Short answer", "panic_note": "I am scared and keep forgetting definitions.", "known_material": "mitosis, meiosis, chromosomes, cytokinesis, cell cycle checkpoints", "...
{ "rescue_plan_markdown": "### Rescue plan\n\n4-step survival plan:\n1. Focus on key definitions: Spend the first 15 minutes reviewing the definitions of mitosis, meiosis, chromosomes, cytokinesis, and cell cycle checkpoints. Write them down clearly and memorize them.\n2. Practice with worked examples: Use the next 1...
{ "model_was_called": true, "contains_required_topic": true, "has_rescue_plan": true, "has_drill_deck": true, "has_triage_clock": true, "has_final_sheet": true, "has_study_receipt": true, "has_field_note_prompt": true, "has_no_hidden_reasoning_tags": true }
{ "real_user_data": false, "sample_disclosed": true, "contains_private_tokens": false, "note": "Public QA sample produced by calling the live Space's MiniCPM endpoint on a disclosed sample scenario, not a real user." }
exam-panic-live-02
live_model_run
2026-06-08T00:00:00
Exam Panic Rescue Gradio Space (live generate endpoint)
https://huggingface.co/spaces/build-small-hackathon/exam-panic-rescue
zerogpu_minicpm_live
Generated with openbmb/MiniCPM4.1-8B (8B) on CUDA/ZeroGPU. Drill source: MiniCPM-generated drills.
{ "scenario_name": "physics numericals", "student_name": "Aarav", "subject": "Physics: work, energy, and power", "time_left_minutes": 120, "exam_format": "Mixed", "panic_note": "I go blank in numericals and forget which formula to use.", "known_material": "work-energy theorem, kinetic energy, potential en...
{ "rescue_plan_markdown": "### Rescue plan\n\n4-step survival plan:\n1. Focus on recall: Spend the first 30 minutes reviewing key formulas and concepts related to work, energy, and power. Write down the work-energy theorem, kinetic energy formula, potential energy formula, and the definition of power.\n2. Worked exam...
{ "model_was_called": true, "contains_required_topic": true, "has_rescue_plan": true, "has_drill_deck": true, "has_triage_clock": true, "has_final_sheet": true, "has_study_receipt": true, "has_field_note_prompt": true, "has_no_hidden_reasoning_tags": true }
{ "real_user_data": false, "sample_disclosed": true, "contains_private_tokens": false, "note": "Public QA sample produced by calling the live Space's MiniCPM endpoint on a disclosed sample scenario, not a real user." }
exam-panic-live-03
live_model_run
2026-06-08T00:00:00
Exam Panic Rescue Gradio Space (live generate endpoint)
https://huggingface.co/spaces/build-small-hackathon/exam-panic-rescue
zerogpu_minicpm_live
Generated with openbmb/MiniCPM4.1-8B (8B) on CUDA/ZeroGPU. Drill source: MiniCPM-generated drills.
{ "scenario_name": "history long answers", "student_name": "Zoya", "subject": "History: nationalism in India", "time_left_minutes": 1440, "exam_format": "Long answer", "panic_note": "I know the chapters but my long answers become messy.", "known_material": "non-cooperation movement, civil disobedience, sa...
{ "rescue_plan_markdown": "### Rescue plan\n\n4-step survival plan:\n1. Start with a clear introduction that outlines the main topic and its importance.\n2. Use the reusable outlines you created to structure your answers, ensuring each section covers key points.\n3. Focus on connecting the dots between the topics, sh...
{ "model_was_called": true, "contains_required_topic": true, "has_rescue_plan": true, "has_drill_deck": true, "has_triage_clock": true, "has_final_sheet": true, "has_study_receipt": true, "has_field_note_prompt": true, "has_no_hidden_reasoning_tags": true }
{ "real_user_data": false, "sample_disclosed": true, "contains_private_tokens": false, "note": "Public QA sample produced by calling the live Space's MiniCPM endpoint on a disclosed sample scenario, not a real user." }
exam-panic-live-04
live_model_run
2026-06-08T00:00:00
Exam Panic Rescue Gradio Space (live generate endpoint)
https://huggingface.co/spaces/build-small-hackathon/exam-panic-rescue
zerogpu_minicpm_live
Generated with openbmb/MiniCPM4.1-8B (8B) on CUDA/ZeroGPU. Drill source: MiniCPM-generated drills.
{ "scenario_name": "math traps", "student_name": "Kabir", "subject": "Math: quadratic equations", "time_left_minutes": 360, "exam_format": "Multiple choice", "panic_note": "MCQ options trick me and I rush the formula.", "known_material": "factorization, quadratic formula, discriminant, completing the squa...
{ "rescue_plan_markdown": "### Rescue plan\n\n4-step survival plan:\n1. Start with the quadratic formula: a, b, c, and the formula itself.\n2. Practice factoring simple quadratics to build confidence.\n3. Focus on recognizing patterns in the options for multiple-choice questions.\n4. Take deep breaths and pace yourse...
{ "model_was_called": true, "contains_required_topic": true, "has_rescue_plan": true, "has_drill_deck": true, "has_triage_clock": true, "has_final_sheet": true, "has_study_receipt": true, "has_field_note_prompt": true, "has_no_hidden_reasoning_tags": true }
{ "real_user_data": false, "sample_disclosed": true, "contains_private_tokens": false, "note": "Public QA sample produced by calling the live Space's MiniCPM endpoint on a disclosed sample scenario, not a real user." }

Exam Panic Rescue App Traces

Public-safe traces for the Build Small Hackathon project Exam Panic Rescue (Backyard AI track).

This dataset supports the Sharing is Caring path and documents the Backyard AI real-user evidence. It focuses on what the app does, not on private build strategy notes.

Configs / files

  • real_user_traces.jsonl (config: real_user) — a real student's anonymized session. A final-year university Machine Learning student (published as "R." with consent) used the live app the day before their exam. The standout outcome: one of the model-written drills closely matched a question that appeared on the actual exam. Real name and identifying details are withheld.
  • app_traces_public.jsonl (config: app_traces) — four full public QA traces with input, generated app outputs, validation flags, runtime notes, and privacy labels. Labeled public QA samples, not real-user outcome data.
  • build_trace_context.jsonl (config: build_context) — small secondary build-context log.
  • APP_TRACE_SCHEMA.md — field definitions for the app trace rows.

Privacy

The real-user trace is published only in anonymized form, with the student's consent. It avoids private notes, tokens, personal contact details, and hidden model reasoning. The QA traces are clearly labeled as samples, not real-user outcomes.

Dataset viewer

This card declares explicit configs so the Hugging Face dataset viewer renders each file as its own selectable view (real_user, app_traces, build_context).

Links

Downloads last month
40

Space using build-small-hackathon/exam-panic-rescue-build-trace 1