datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
R2E-Gym-Subset-Verified
R2E-Gym-Subset-Verified
Gold-patch-validated subset of
R2E-Gym/R2E-Gym-Subset
(paper). The train split contains
4,522 / 4,578 rows (98.78%) verified scoreable end-to-end: apply the gold patch, run the
upstream /testbed/run_tests.sh baked into the row's image, check the parsed outcomes against
expected_output_json.
Changes vs upstream
Validation-only subset — our passes, run in fresh sandboxes per row: one full pass at
concurrency 200, then a 10× retry pass over… See the full description on the dataset page: https://huggingface.co/datasets/PrimeIntellect/R2E-Gym-Subset-Verified.R2E-Gym-Full
R2E-Gym Subset Filtered for MAGRPO
Filtered subset of R2E-Gym optimized for 2-agent MAGRPO training with 7B models.
Dataset Statistics
Total instances: 167
Format: Issue description + Oracle files in prompt
Optimized for: 2-agent collaboration, 7B models
Filtering Criteria (SWE-bench Lite Style)
Problem statement: >40 words (up to 500 for context window)
Must have non-empty oracle patch (non-test file changes)
File count: Exactly 1 oracle file (single-file… See the full description on the dataset page: https://huggingface.co/datasets/ryankamiri/R2E-Gym-Full.R2-Bench
R2-Bench
R2-Bench is a benchmark dataset for evaluating LLM routing with joint model and token budget optimization. It contains 30,968 queries evaluated across 10 LLMs at 16 token budget levels, with LLM-judge quality scores.
Associated with R2-Router (code), under review at ICML 2026.
Dataset Structure
data/
├── meta-llama/
│ ├── Llama-3.1-70B-Instruct/
│ │ ├── 10_judge.csv
│ │ ├── 20_judge.csv
│ │ ├── ...
│ │ └── 8000_judge.csv
│ └──… See the full description on the dataset page: https://huggingface.co/datasets/JiaqiXue/R2-Bench.R2E-Gym
R2E-Gym: OpenCode, Codex, and Claude Code environments
This repository contains three environment variants of
R2E-Gym/R2E-Gym-Subset.
Each variant contains the same 4,578 tasks in five Parquet shards, with the
original 14-column schema and task order. Only docker_image is replaced with
the corresponding public image containing the selected agent.
Configuration
Files
Docker image prefix
opencode
opencode/data/*.parquet
docker.io/loongsage/r2e-gym:opencode_
codex… See the full description on the dataset page: https://huggingface.co/datasets/loongsage/R2E-Gym.task1386_anli_r2_entailment
Dataset Card for Natural Instructions (https://github.com/allenai/natural-instructions) Task: task1386_anli_r2_entailment
Additional Information
Citation Information
The following paper introduces the corpus in detail. If you use the corpus in published work, please cite it:
@misc{wang2022supernaturalinstructionsgeneralizationdeclarativeinstructions,
title={Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Tasks}… See the full description on the dataset page: https://huggingface.co/datasets/Lots-of-LoRAs/task1386_anli_r2_entailment.R2E-Gym-Subset
R2E-Gym Subset Filtered for MAGRPO
Filtered subset of R2E-Gym optimized for 2-agent MAGRPO training with 7B models.
Dataset Statistics
Total instances: 54
Format: Issue description + Oracle files in prompt
Optimized for: 2-agent collaboration, 7B models
Filtering Criteria (SWE-bench Lite Style)
Problem statement: >40 words (up to 500 for context window)
Must have non-empty oracle patch (non-test file changes)
File count: Exactly 1 oracle file (single-file… See the full description on the dataset page: https://huggingface.co/datasets/ryankamiri/R2E-Gym-Subset.filtered-R2EGym-SFT-Trajectories
Dataset Summary
This dataset contains 3,218 high-quality agent trajectories where a programming agent interacts with a repository environment to diagnose and fix GitHub issues. Each sample is a multi-turn conversation following the standard chat format.
Statistics
Metric
Value
Total samples
3,218
Avg turns/sample
33.2
Min turns
15
Max turns
85
Use Cases
SFT fine-tuning: Train code agents to resolve real-world GitHub issues… See the full description on the dataset page: https://huggingface.co/datasets/ubicloud/filtered-R2EGym-SFT-Trajectories.r20-portfolio-ai-perception
Portfolio Interference in LLM Brand Perception (R20 to R21)
Supersession note: This dataset originally backed R20 (2026ab, superseded). R21 (2026ac, DOI 10.5281/zenodo.19765401) supersedes both R8 (2026q) and R20. R21 merges R8 theory with R20 empirical (9,925 obs across 40 brands, 13 models, 7 traditions) into a single analytical-empirical paper. New citations should reference Zharnikov (2026ac).
Dataset DOI: 10.57967/hf/8380
Current Paper (R21): 10.5281/zenodo.19765401 --… See the full description on the dataset page: https://huggingface.co/datasets/spectralbranding/r20-portfolio-ai-perception.coyote-r2-training-data
coyote-r2-training-data
Coyote the Coder — training round R2. Instruction/tool-call data (system+user+assistant message triples with task labels) used to fine-tune the Coyote coder model (Qwen3.5-4B-Base).
Contents
train.jsonl (348 rows)
validation.jsonl (61 rows)
Format
JSON Lines (.jsonl), one example per line.
Provenance
Original content authored for the Coyote the Coder project (Michael Anthony Falabella).
R2R_queryThis is the training and validation query set used by the paper R2R: Efficiently Navigating Divergent Reasoning Paths with Small-Large Model Token Routing. This dataset contains token-level routing labels generated to train a lightweight router that selectively uses a Large Language Model (LLM) for critical, path-divergent tokens during inference, improving efficiency without sacrificing accuracy.
Roads to Rome (R2R) is a neural token router that efficiently combines Large Language Models… See the full description on the dataset page: https://huggingface.co/datasets/nics-efc/R2R_query.r2ata
Dataset Card for Reasoning Robustness to Adversarial Typo Attacks (R2ATA)
Check out our paper here
Check out our website here
Dataset Overview
This dataset comprises adversarially edited questions derived from three well-known benchmark datasets: BBH, MMLU, and GSM8K. These questions have been specifically modified to evaluate and challenge the robustness of LLMs by introducing typographical errors.
Source Datasets
BBH (Big-Bench Hard):
A… See the full description on the dataset page: https://huggingface.co/datasets/project-works/r2ata.R2E-Gym-Collab-v3
R2E-Gym-Collab-v3
A high-quality subset of R2E-Gym optimized for multi-agent RL training (MAGRPO).
What's Special
This dataset is filtered for optimal gradient signal in Level 3 (test execution) rewards:
2-20 failing tests per instance (good gradient signal)
Collaboration-suitable (AI-verified two-agent task decomposition)
Test failures in prompt (explicit error context for the model)
Why This Filtering?
Test Count
L3 Reward per Fix
Problem
1… See the full description on the dataset page: https://huggingface.co/datasets/ryankamiri/R2E-Gym-Collab-v3.R2E-Gym-Collab-v2
R2E-Gym-Collab-v2
This dataset extends ryankamiri/R2E-Gym-Collab by adding test failure information to each prompt.
What's New
Each prompt now includes a [TEST_FAILURES] section showing:
The actual pytest output from failing tests (before the fix)
Error messages, assertion failures, and tracebacks
This gives models context about WHAT needs to be fixed, not just WHERE
Prompt Structure
[ISSUE]
Bug description, expected vs actual behavior
[/ISSUE]… See the full description on the dataset page: https://huggingface.co/datasets/ryankamiri/R2E-Gym-Collab-v2.