datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
openswe-tasks-patched-v5static-analysis-evalA dataset of 76 Python programs taken from real Python open source projects (top 100 on GitHub),
where each program is a file that has exactly 1 vulnerability as detected by a particular static analyzer (Semgrep), used in the paper Patched MOA: optimizing inference for diverse software development tasks.
OpenAI used the synth-vuln-fixes and fine-tuned
a new version of gpt-4o is now the SOTA on this benchmark. More details and code is available from their repo.
More details on the benchmark… See the full description on the dataset page: https://huggingface.co/datasets/patched-codes/static-analysis-eval.cc12m_openai_clip-vit-base-patch32_image_image_retrieval_pairs_2022-09-13vulnerability-cwe-patch
Description
This dataset, CIRCL/vulnerability-cwe-patch, provides structured, real-world vulnerabilities enriched with CWE identifiers and corresponding patches from platforms like GitHub and GitLab. It is designed to support the development of tools for vulnerability classification, triage, and automated remediation. Each entry includes metadata such as CVE/GHSA ID, a description, CWE categorization, and links to verified patch commits with associated diff content and commit… See the full description on the dataset page: https://huggingface.co/datasets/CIRCL/vulnerability-cwe-patch.generate-readme-eval
Generate README Eval
The generate-readme-eval is a dataset (train split) and benchmark (test split) to evaluate the effectiveness of LLMs
when summarizing entire GitHub repos in form of a README.md file. The datset is curated from top 400 real Python repositories
from GitHub with at least 1000 stars and 100 forks. The script used to generate the dataset can be found here.
For the dataset we restrict ourselves to GH repositories that are less than 100k tokens in size to allow us to… See the full description on the dataset page: https://huggingface.co/datasets/patched-codes/generate-readme-eval.patchlet-embed-preprocessedswe_gym_annotate_with_patchLLaVa-Instruct-150K-clip-vit-base-patch32SynthMat_Patches_DSgithub-patchesswe_rebench_patchedrl__24GPU_base__swe_rebench_patched_oracle__r2egym-nl2bash-stackR2E-TestgenAgent-PatchesR2EGym-VerifierTrajectories-PatchOnlygithub-patches-genesysimport re
import json
from datasets import load_dataset
PROMPT_TEMPLATE = """\
We are currently solving the following issue within our repository. Here is the issue text:
--- BEGIN ISSUE ---
{issue}
--- END ISSUE ---
Below are some code segments, each from a relevant file. One or more of these files may contain bugs.
--- BEGIN FILES ---
{file_context}
--- END FILES ---
Please first localize the bug based on the issue statement, and then generate a patch according to the `git diff` format… See the full description on the dataset page: https://huggingface.co/datasets/rasdani/github-patches-genesys.PatchBench
PatchBench
PatchBench is a benchmark for evaluating AI agents on realistic vulnerability patching tasks: 213 tasks drawn from 32 popular GitHub C/C++ projects. It selects vulnerabilities whose ground-truth fixes lie outside the crash stack, and uses vulnerability transplant plus code mutation to mitigate surface-level fixes and patch memorization.
This repository holds the task metadata, one row per task to identify the project, the exact repository state, the crash, and the… See the full description on the dataset page: https://huggingface.co/datasets/ai-sec-lab/PatchBench.mtg-scryfall-cropped-art-embeddings-siglip-so400m-patch14-384cc12m_openai_clip-vit-base-patch32_image_image_retrieval_pairs_2022-09-15rl__24GPU_shaped__swe_rebench_patched_oracle__r2egym-nl2bash-stackmscoco_train_2014_openai_clip-vit-base-patch32_image_image_retrieval_pairs_2022-09-13wikitext-103-raw-v1_sents_min_len10_max_len30_openai_clip-vit-base-patch32Patchnoisseur
Patchnoisseur
A connoisseur's cellar of CVEs: every NVD CVE joined to its fixing-commit
diff (when one could be found), its NVD description, and its
associated CWE(s) (id, name, short description) — served as a single
Parquet dataset.
351 884 CVEs · 25 015 with a real git diff attached · CVE-1999 → CVE-2026
· ~744 MB on disk (zstd-compressed Parquet, sharded ~300 MB each).
What's in it
One row per CVE in the NVD feed. CVEs without a retrievable patch are… See the full description on the dataset page: https://huggingface.co/datasets/michoo42/Patchnoisseur.a3-rl-DCAgent_r2egym-patched-full-oracleswe_rebench_patched_oraclegithub-patches-decontaminated# removed all repos of SWE-bench and RepoBench
repos = [
"astropy",
"django",
"flask",
"matplotlib",
"seaborn",
"requests",
"xarray",
"pylint",
"pytest",
"scikit-learn",
"sphinx",
"sympy",
]
act_patch_llama_3.1_8b_counterfact
Training Language Models to Explain Their Own Computations
Paper | Code
This dataset contains activation patching results used for training explainer models to predict how internal interventions affect target model outputs. It was introduced in the paper "Training Language Models to Explain Their Own Computations".
Dataset Summary
The dataset covers the Activation Patching task for the Llama-3.1-8B target model, where explainer models learn to predict the effects of… See the full description on the dataset page: https://huggingface.co/datasets/Transluce/act_patch_llama_3.1_8b_counterfact.r2egym-patched-full-oracle-qwen3.5-122b-131k-opencode-literal-rescue-traces
Agent trace dataset
Decoding the literal token IDs
The prompt_token_ids / completion_token_ids / logprobs columns are the
verbatim tokens the serving engine emitted, stored PER AGENT STEP as a
list-of-lists (one inner list per turn). To turn them back into text you MUST
use the exact tokenizer the model was served with — a generic same-family
tokenizer will decode word tokens to garbage.
Served model / tokenizer source: Qwen/Qwen3.5-122B-A10B-FP8
from transformers… See the full description on the dataset page: https://huggingface.co/datasets/open-athena/r2egym-patched-full-oracle-qwen3.5-122b-131k-opencode-literal-rescue-traces.Kimi-2.5-swe_rebench_patched-maxeps-32kbln600-img-patch
BLN600 Image Patches
This dataset provides BLN600's image patches for fine-tuning vision-language models on post-OCR correction, introduced in "Image-Informed Post-OCR Correction with Vision-Language Models" (EMNLP 2026 Findings). Each patch corresponds to a text sequence in BLN600 and is cropped from the Gale British Library Newspapers collection, using word-level bounding boxes from the collection's ALTO XML OCR layout data. It is intended to be used alongside the code and… See the full description on the dataset page: https://huggingface.co/datasets/pykale/bln600-img-patch.patch_tasks_vllm
Dataset Card for Patch-Based Visual Question Answering Dataset
Dataset Details
Dataset Description
This dataset contains approximately 305,000 triplets of question, answer, and image designed for patch-based visual reasoning tasks.
A standard question in this dataset is formatted as follows:
Image Grid: The image is divided into a 4x4 grid of 16 equal-sized patches. Patches are numbered sequentially from the top-left corner and moving right, then down to the… See the full description on the dataset page: https://huggingface.co/datasets/yurkes/patch_tasks_vllm.