datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
DeepSWEGym2
Dataset Description
This dataset is a filtered and deduplicated version of a merge containing many high quality SWE datasets, it aims to improve benchmark results on DeepSWE-style problems, benchmarks, and general coding skills.
It is specifically filtered for rows with complex/long code problems in the original datasets, having an average row size of 214.19kb, a total uncompressed size of 17.56GB, and a total of 85974 examples.
Dataset Details
Curated by:… See the full description on the dataset page: https://huggingface.co/datasets/MoreThought/DeepSWEGym2.DeepSWEGym2-Ultra
Dataset Description
This dataset is an EXTREMELY filtered and deduplicated version of DeepSWE-Gym2-Edu, it aims to improve benchmark results on DeepSWE-style problems, benchmarks, and general coding skills.
It is specifically filtered for rows with complex/long code problems in the original datasets, having an average row size of 8.27MB, a total uncompressed size of 8.27GB, and a total of 1000 examples.
Dataset Details
Curated by: MoreThought
Funded by:… See the full description on the dataset page: https://huggingface.co/datasets/MoreThought/DeepSWEGym2-Ultra.deep-swe
DeepSWE
DeepSWE is a benchmark for measuring frontier coding agents on original, long-horizon software engineering tasks drawn from active open-source repositories. The benchmark includes 113 tasks across TypeScript, Go, Python, JavaScript, and Rust, with isolated environments and program-based verifiers.
Task format
DeepSWE tasks use the Harbor task format:
task.toml Metadata: repository, base commit, language, prebuilt image, resource limits… See the full description on the dataset page: https://huggingface.co/datasets/datacurve/deep-swe.DeepSWEGym2-Edu
Dataset Description
This dataset is a filtered and deduplicated version of a merge containing many high quality SWE datasets, it aims to improve benchmark results on DeepSWE-style problems, benchmarks, and general coding skills.
It is specifically filtered for rows with complex/long code problems in the original datasets, having an average row size of 291.74kb, a total uncompressed size of 14.93GB, and a total of 53649 examples.
Dataset Details
Curated by:… See the full description on the dataset page: https://huggingface.co/datasets/MoreThought/DeepSWEGym2-Edu.deep-swe-1-1-materialized
DeepSWE 1.1 — materialized
A tabular materialization of DeepSWE
v1.1 — Datacurve's 113-task benchmark for coding agents — repackaged from
datacurve-ai/deep-swe into one
parquet row per task. This is a third-party repack for tooling convenience,
not an official Datacurve release.
Source commit: see manifest.json (source_commit) — every file is
carried over unmodified into columns.
Integrity: manifest.json records the parquet's sha256 and a per-task
content hash (sha256 over each… See the full description on the dataset page: https://huggingface.co/datasets/luolc/deep-swe-1-1-materialized.DeepSWEGym2-Full
Dataset Description
This dataset is a merge containing many high quality SWE datasets, it aims to improve benchmark results on DeepSWE-style problems, benchmarks, and general coding skills.
It is NOT specifically filtered for rows with complex/long code problems in the original datasets, despite still having an average row size of 169.08kb, a total uncompressed size of 19.80GB, and a total of 122791 examples.
Dataset Details
Curated by: MoreThought
Funded by:… See the full description on the dataset page: https://huggingface.co/datasets/MoreThought/DeepSWEGym2-Full.DeepSWEGym-Edu
Dataset Description
This dataset is a heavily filtered version of all the SWE-bench/SWE-smith-lang datasets (expect php) merged together (originally 88k rows total), it aims to improve benchmark results on DeepSWE-style problems, benchmarks, and general coding skills.
It is specifically filtered for rows with very complex/long code problems in the original datasets, having an average row size of 101.81kb, a total uncompressed size of 4.29GB, and 42529 examples total.… See the full description on the dataset page: https://huggingface.co/datasets/MoreThought/DeepSWEGym-Edu.DeepSWEGym-Full
Dataset Description
This dataset is a merged version of all the SWE-bench/SWE-smith-lang datasets (88k rows total), it aims to improve benchmark results on DeepSWE-style problems, benchmarks, and general coding skills.
It is NOT specifically filtered for rows with complex/long code problems in the original datasets, despite still having an average row size of 85.4kb, a total uncompressed size of 7.53GB, and 88130 examples total.
Dataset Details
Curated by:… See the full description on the dataset page: https://huggingface.co/datasets/MoreThought/DeepSWEGym-Full.DeepSWE-Agent-Kimi-K2-Trajectories-2.8Kdeepswe-verifier-2582-v1deepswe-mini
deepswe-mini
16 of the 113 tasks in DeepSWE v1.1, picked so that running just these ranks models the same way the full benchmark does.
DeepSWE is a good benchmark and an expensive one. Every task is a long-horizon feature request in its own container, and a full pass takes close to two days of agent time run one task at a time. If you are comparing models, agent harnesses or prompts, and the differences you care about are more than a few points, these 16 tasks give you the same… See the full description on the dataset page: https://huggingface.co/datasets/LocalLLaMA/deepswe-mini.DeepSWE-Agent-Kimi-K2-Trajectories-Rejection-Samplingbirdcode-deepswe-k1d
BirdCode on DeepSWE — k=1, single attempt, no web tools
⚠️ Reading the metric correctly: The summary card's "Average f2p 0.89" is the
test-case-level pass fraction (f2p_passed/f2p_total, averaged per task) — it is
NOT the official DeepSWE leaderboard metric. The official binary score is the
reward field (1 only when all F2P and P2P tests pass): 60/113 = 0.531.
Per-trial reward values are visible in each trial's rewards block below.
Evaluation of BirdCode (a from-scratch… See the full description on the dataset page: https://huggingface.co/datasets/tangchen-ai/birdcode-deepswe-k1d.deepswe-verifier-merged-with-regression-with-filenamesdeepswe-verifier-2582-v2deepswe-verifier-merged-with-regressiondeepswe-verifier-merged-with-regression-with-filenames-cleandeepswe-verifier-3582-exitreason-agent-priority-v1deepswe-swebv-eval-n16-verifier-v1deepswe-verifier-merged-with-regression-with-filenames-and-rewards-v2deepswe-verifier-only-matching-pairs-v1deepswe-verifier-merged-with-regression-with-filenames-and-rewardsdeepswe-prm-embeddings-8k
DeepSWE PRM evaluation embeddings (Qwen3-8B, 8k)
Qwen3-8B last-token-pooled embeddings (4096-d) of every step of the DeepSWE evaluation
rollouts (mini-swe-agent, Claude-Opus-5, 4 rollouts per task): 44,409 steps, 449 rollouts,
113 tasks. Embedded with preprocessing/deepswe/embed_shard.py at max_model_len 8192.
column
description
trajectory_id
rollout id
step_idx
step order within the rollout
task_id
DeepSWE task
model, config
policy model / rollout config… See the full description on the dataset page: https://huggingface.co/datasets/tarsur385/deepswe-prm-embeddings-8k.deepswe-verifier-exitreason-agent-priority-v1deepswe-swebv-eval-n16-verifier-v1-with-regressiondev_set_v2_DeepSWE_Preview_20260501_185747deepswe-prm-train-embeddings-8k
DeepSWE PRM training embeddings (Qwen3-8B, 8k)
Frozen Qwen3-8B last-token-pooled embeddings (4096-d, float16) of every step of the
DeepSWE training-pool rollouts: 405,919 steps · 4,701 trajectories · 113 tasks.
This is the data the released DeepSWE PRM heads
(tarsur385/deepswe-prm-heads-8k) were fine-tuned on.
Embedded with preprocessing/deepswe/embed_shard.py at max_model_len 8192: the state is the
chat-templated step context truncated to its last 8191 tokens; the action is the… See the full description on the dataset page: https://huggingface.co/datasets/tarsur385/deepswe-prm-train-embeddings-8k.deepswe-verifier-debug-condense-v1dev_set_v2_rl__r2egym_deepswe_fp8_terminus_2_32b_20260406_202950swebench_verified_random_100_folders_DeepSWE_Preview_20260501_185848