Datasets:

reasoning-degeneration-dev
/

gepa-exp-contrastive-vanilla-20260220-083526

run_id stringclasses 1 value	iteration int64 1 6	cumulative_tokens int64 9.56k 51.6k	best_val_score float64 0.18 0.31	accepted bool 2 classes	candidate_idx int64 0 2	candidate_prompt stringclasses 3 values	best_prompt stringclasses 2 values
fixed_iter_vanilla_k3	1	9,558	0.177778	false	0	You are a helpful assistant. You are given a question and you need to answer it. The answer should be given at the end of your response in exactly the format '### <final answer>'	You are a helpful assistant. You are given a question and you need to answer it. The answer should be given at the end of your response in exactly the format '### <final answer>'
fixed_iter_vanilla_k3	2	21,686	0.311111	true	1	Follow these guidelines to avoid the errors seen in the failed attempts and to match the successful behavior: 1) Final-answer formatting: - End your response with exactly one line in the format: "### <final answer>" (e.g., "### 242"). - Do not add any extra tags (like <final answer></final answer>), boxes, or additio...	Follow these guidelines to avoid the errors seen in the failed attempts and to match the successful behavior: 1) Final-answer formatting: - End your response with exactly one line in the format: "### <final answer>" (e.g., "### 242"). - Do not add any extra tags (like <final answer></final answer>), boxes, or additio...
fixed_iter_vanilla_k3	3	21,686	0.311111	false	1	Follow these guidelines to avoid the errors seen in the failed attempts and to match the successful behavior: 1) Final-answer formatting: - End your response with exactly one line in the format: "### <final answer>" (e.g., "### 242"). - Do not add any extra tags (like <final answer></final answer>), boxes, or additio...	Follow these guidelines to avoid the errors seen in the failed attempts and to match the successful behavior: 1) Final-answer formatting: - End your response with exactly one line in the format: "### <final answer>" (e.g., "### 242"). - Do not add any extra tags (like <final answer></final answer>), boxes, or additio...
fixed_iter_vanilla_k3	4	30,263	0.311111	false	1	Follow these guidelines to avoid the errors seen in the failed attempts and to match the successful behavior: 1) Final-answer formatting: - End your response with exactly one line in the format: "### <final answer>" (e.g., "### 242"). - Do not add any extra tags (like <final answer></final answer>), boxes, or additio...	Follow these guidelines to avoid the errors seen in the failed attempts and to match the successful behavior: 1) Final-answer formatting: - End your response with exactly one line in the format: "### <final answer>" (e.g., "### 242"). - Do not add any extra tags (like <final answer></final answer>), boxes, or additio...
fixed_iter_vanilla_k3	5	42,767	0.311111	true	2	You are a helpful assistant. You are given a question and you need to answer it. The answer should be given at the end of your response in exactly the format '### <final answer>'. To avoid the specific errors seen in the failed attempts, follow these rules derived from where the successful solutions differed: General...	Follow these guidelines to avoid the errors seen in the failed attempts and to match the successful behavior: 1) Final-answer formatting: - End your response with exactly one line in the format: "### <final answer>" (e.g., "### 242"). - Do not add any extra tags (like <final answer></final answer>), boxes, or additio...
fixed_iter_vanilla_k3	6	51,600	0.311111	false	2	You are a helpful assistant. You are given a question and you need to answer it. The answer should be given at the end of your response in exactly the format '### <final answer>'. To avoid the specific errors seen in the failed attempts, follow these rules derived from where the successful solutions differed: General...	Follow these guidelines to avoid the errors seen in the failed attempts and to match the successful behavior: 1) Final-answer formatting: - End your response with exactly one line in the format: "### <final answer>" (e.g., "### 242"). - Do not add any extra tags (like <final answer></final answer>), boxes, or additio...

gepa-exp-contrastive-vanilla-20260220-083526

GEPA prompt optimization experiment on AIME math problems.

Task LM: openai/gpt-4.1-mini | Reflection LM: openai/gpt-5 | Reflection Mode: contrastive | Last updated: 2026-02-20 18:21 UTC

Results

Run	Method	k	Mode	Val Score	Test Acc	Tokens	Cost	Time
fixed_vanilla_k3	vanilla	3	contrastive	31.11%	44.00%	51,600	$0.1236	32317s

Learning Curves

Experiment Config

{
  "script_name": "run_experiment.py",
  "model": "openai/gpt-4.1-mini",
  "reflection_lm": "openai/gpt-5",
  "reflection_mode": "contrastive",
  "hyperparameters": {
    "task_lm": "openai/gpt-4.1-mini",
    "reflection_lm": "openai/gpt-5",
    "reflection_mode": "contrastive",
    "seed_prompt": "You are a helpful assistant. You are given a question and you need to answer it. The answer should b..."
  },
  "input_datasets": [
    "AI-MO/aimo-validation-aime",
    "MathArena/aime_2025"
  ],
  "description": "GEPA prompt optimization experiment on AIME (reflection_mode=contrastive)",
  "num_runs": 1,
  "last_updated": "2026-02-20 18:21 UTC"
}

Dataset Configs

This repo contains multiple configs (subsets). Load them with:

from datasets import load_dataset

results    = load_dataset("reasoning-degeneration-dev/gepa-exp-contrastive-vanilla-20260220-083526", "results", split="train")
traces     = load_dataset("reasoning-degeneration-dev/gepa-exp-contrastive-vanilla-20260220-083526", "traces", split="train")
val_traces = load_dataset("reasoning-degeneration-dev/gepa-exp-contrastive-vanilla-20260220-083526", "val_traces", split="train")
curves     = load_dataset("reasoning-degeneration-dev/gepa-exp-contrastive-vanilla-20260220-083526", "curves", split="train")
rlm_call_traces = load_dataset("reasoning-degeneration-dev/gepa-exp-contrastive-vanilla-20260220-083526", "rlm_call_traces", split="train")

Config	Description
`results`	Per-run summary: method, k, val/test scores, tokens, cost, optimized prompt
`traces`	Per-example test set traces with model responses and scores
`val_traces`	Per-iteration validation traces with the prompt used at each step
`curves`	Per-iteration learning curves: tokens, scores, candidate prompts
`rlm_call_traces`	Hierarchical LLM call traces: every RLM iteration, sub-call, prompt, and response

State file: state.json

For experiment design details, see EXPERIMENTS-DESIGN.md

Downloads last month: 35