datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
rlvr-reward-hacking-scale-no-conftest-20260909-completion
Matched no-conftest RLVR study 20260909-completion
Lossless research records, grouped by model and trajectory type. Only the listed
configurations have published records. Canary diagnostics are excluded from study
estimates; run status in provenance distinguishes retired diagnostics from active
or completed training. Valid failures, refusals and truncations are retained.
The train split name is a dataset-loader convention; record_type identifies
whether a record is training… See the full description on the dataset page: https://huggingface.co/datasets/lucabaroni/rlvr-reward-hacking-scale-no-conftest-20260909-completion.hack-ignition-benchmark
hack-ignition benchmark — data, v0.1.6
Training trajectories of reinforcement-learning runs on exploitable graders, for studying and predicting when RL
comes to produce exploits. Each family is a set of GRPO runs over configurations of (start model, prompt,
training set, grader / reward structure, recipe), with one or more seeds per configuration. Every family stores
what its training logs contain — per-step exploit, task and reward rates, the item × step exploit record… See the full description on the dataset page: https://huggingface.co/datasets/EleutherAI/hack-ignition-benchmark.rlvr-reward-hacking-scale-no-conftest-20260909
Matched no-conftest RLVR study 20260909
Complete immutable training, monitoring and comparison trajectories for six models.
All valid outcomes are retained, including refusals, failures and truncations.
The train split name is a dataset-loader convention; record_type identifies
whether a record is training, monitoring, comparison, or a derived judgment.
import json
from datasets import load_dataset
rows = load_dataset("lucabaroni/rlvr-reward-hacking-scale-no-conftest-20260909"… See the full description on the dataset page: https://huggingface.co/datasets/lucabaroni/rlvr-reward-hacking-scale-no-conftest-20260909.HackerSignal
HackerSignal
A large-scale, multi-source dataset linking hacker community discourse, exploit databases, vulnerability advisories, and fix commits through a shared CVE identifier space.
Overview
Statistic
Value
Documents
7,447,646 (exact-deduplicated)
Sources
64 public forum/source identifiers
Source layers
8
Temporal span
1988--2026
CVE-linked rows
360,004
Benchmark tasks
3
Quick Start
from datasets import load_dataset
# Load the… See the full description on the dataset page: https://huggingface.co/datasets/DatasetSubmission/HackerSignal.hacktheboxhackernews-comments
Hackernews Comments Dataset
A dataset of all HN API items from id=0 till id=41422887 (so from 2006 till 02 Sep 2024). The dataset is build by scraping the HN API according to its official schema and docs. Scraper code is also available on github: nixiesearch/hnscrape
Dataset contents
No cleaning, validation or filtering was performed. The resulting data files are raw JSON API response dumps in zstd-compressed JSONL files. An example payload:
{
"by": "goldfish"… See the full description on the dataset page: https://huggingface.co/datasets/nixiesearch/hackernews-comments.syntheticagenda-parser-tool-traces
Agenda Parser — tool-calling reasoning traces
ReAct tool-calling traces for the Agenda Parser
agents: each row is one agent step — a {system, user, assistant} chat example
where the assistant emits a single JSON action {"thought", "tool", "args"}.
Two agents are covered (tagged by meta.domain):
agenda — the uploaded-packet research agent, over real public-meeting agenda
packets (tools: list/read items, semantic + exact search, summarize, report).
Each agenda row's meta.unit_id… See the full description on the dataset page: https://huggingface.co/datasets/build-small-hackathon/agenda-parser-tool-traces.alchemist-shell.ai-hackathon-2025This project is described in detail at this website:
https://alchemist-shellai-hackathon-2025.readthedocs.io/en/latest/
The codes and relevant materials are available here:
https://github.com/Sukantabasu/alchemist-shell.ai-hackathon-2025
The trained models (in pkl format) are stored in this HF repository.
jawbreaker-scam-defense-data
Jawbreaker Scam Defense Data
Synthetic and sanitized training/eval data for Jawbreaker, a local-first scam defense app for someone you love.
Jawbreaker turns a suspicious text, email, or DM into a plain-English safety card: the risk, the warning signs, and the safest next step before someone replies, clicks, or pays.
Contents
eval/: scam-defense evaluation sets from smoke checks through hard calibration suites.
eval/reports/: guarded evaluation reports for the… See the full description on the dataset page: https://huggingface.co/datasets/build-small-hackathon/jawbreaker-scam-defense-data.science-on-a-sphere-prompt-completions
Dataset Card for Science On a Sphere QA Dataset
Dataset Details
Dataset Description
This dataset comprises question-and-answer (QA) pairs generated from NOAA's Science On a Sphere (SOS) website, including support documentation and the dataset catalog. Each entry contains a prompt and a corresponding completion, designed to support educational and research use cases in Earth science.
This dataset includes a custom dataset_script.py and a consolidated file… See the full description on the dataset page: https://huggingface.co/datasets/HacksHaven/science-on-a-sphere-prompt-completions.readability-es-hackathon-pln-public
Dataset Card for [readability-es-sentences]
Dataset Description
Compilation of short Spanish articles for readability assessment.
Dataset Summary
This dataset is a compilation of short articles from websites dedicated to learn Spanish as a second language. These articles have been compiled from the following sources:
Coh-Metrix-Esp corpus (Quispesaravia, et al., 2016): collection of 100 parallel texts with simple and complex variants in Spanish. These texts… See the full description on the dataset page: https://huggingface.co/datasets/somosnlp-hackathon-2022/readability-es-hackathon-pln-public.hacker-news
Hacker News Dataset
Dataset Summary
This dataset is derived from the official Hacker News data provided via the Hacker News Firebase API. It contains user-generated content including stories, comments, and metadata from the Hacker News platform.
Hacker News is a social news website focusing on computer science, entrepreneurship, and technology. The dataset captures real-world discussions, technical conversations, and community interactions over time.
The data was… See the full description on the dataset page: https://huggingface.co/datasets/TheFinAI/hacker-news.rlvr-reward-hacking-transcripts
RLVR reward-hacking full trajectories
This release contains 900 full held-out trajectories from three policies trained with
reinforcement learning from verifiable rewards (RLVR) in a deliberately vulnerable
CodeContests evaluator: 300 each from the final Qwen3.5-9B, GPT-OSS-120B, and Nemotron-3-Super-120B-A12B
checkpoints. Each row preserves the task, tests, complete prompts, native
reasoning, final answer, rendered and sampled token IDs, token log-probabilities, sampling… See the full description on the dataset page: https://huggingface.co/datasets/lucabaroni/rlvr-reward-hacking-transcripts.hackingai-training
hackingai-training
Training dataset for a specialized offensive-security AI (agentic tool-calling + security Q&A).
Built by merging public security datasets with the user's private corpus ("Bucket"), deduplicated,
formatted as Qwen3 ChatML.
Stats (round 3 — Claude's gap-fix)
train.jsonl: 951,150 samples (3.6 GB) — ChatML, categories:
agentic 303,972 | sft 225,574 | distill 100,608 | cve 54,240 | distill_code 55,600
distill_sec_reason 67,811 | offensive 38,553 |… See the full description on the dataset page: https://huggingface.co/datasets/Panagiotis12/hackingai-training.rlvr-reward-hacking-scale-no-conftest-20260909-budget8192
Matched no-conftest RLVR study 20260909-budget8192
Retired before study training. This dataset contains only validation diagnostics for discarded forced-reasoning and code-prefix policies, including failures and interruptions. No study training or base/50%/final comparison evaluations were launched under those policies. They are excluded from the active completion-reward study. All available diagnostic records are preserved losslessly below.
Lossless research records, grouped by… See the full description on the dataset page: https://huggingface.co/datasets/lucabaroni/rlvr-reward-hacking-scale-no-conftest-20260909-budget8192.hackernews-stories
A HackerNews Stories dataset
This dataset is based on nixiesearch/hackernews-comments dataset:
for each item of type=story we downloaded the target URL. Out of ~3.8M stories ~2.1M are still reachable.
each story HTML was parsed using trafilatura library
we store article text in markdown format along with all page-specific metadata.
Dataset stats
date coverage: xx.2006-09.2024, same as in upstream nixiesearch/hackernews-comments dataset
total scraped pages: 2150271… See the full description on the dataset page: https://huggingface.co/datasets/nixiesearch/hackernews-stories.rlvr-reward-hacking-mid-checkpoint-transcripts
RLVR reward-hacking mid-checkpoint full trajectories
This release contains 600 full held-out trajectories from intermediate RLVR
checkpoints selected to yield substantially more balanced reward-hacking datasets: 300
from Qwen3.5-9B at optimizer update 110 and 300 from GPT-OSS-120B at update 180.
Each row preserves the task and tests, complete prompts, native reasoning, final answer,
rendered and sampled token IDs, token log-probabilities, sampling metadata, extracted
files… See the full description on the dataset page: https://huggingface.co/datasets/lucabaroni/rlvr-reward-hacking-mid-checkpoint-transcripts.Alpaca_Backdoor_Dataset
Alpaca Backdoored Triggered Dataset
Dataset Description
This dataset is a backdoored variant of the Alpaca instruction-following dataset, created for research on backdoor attacks, robustness, and detection in instruction-tuned language models.
A subset of samples contains an intentional trigger pattern inserted in the instruction. When the trigger is present, the model is induced to produce a predefined target behavior, while non-triggered samples remain unchanged.
⚠️… See the full description on the dataset page: https://huggingface.co/datasets/Hackxm/Alpaca_Backdoor_Dataset.kirana-detective-build-traces
Kirana Detective — Claude Code Build Sessions
Raw Claude Code (claude-sonnet-4-6) session traces recorded while building
Kirana Detective AI for the HuggingFace Build Small Hackathon 2026.
Each .jsonl file is one coding session. Together they cover the entire
build — from first commit to final submission.
What's Inside
Sessions
Agent
Coverage
11 JSONL files
Claude Code (Sonnet 4.6)
Full project build
Sessions include
Designing the… See the full description on the dataset page: https://huggingface.co/datasets/build-small-hackathon/kirana-detective-build-traces.hackathon-advisor-codex-traces
Hackathon Advisor Codex Session Traces
Real Codex session logs for the Hackathon Advisor project, selected from local Codex
rollout JSONL files and redacted before publication. The event stream preserves user
requests, assistant messages, tool calls, tool outputs, browser/search events, and
minimal session provenance needed to audit how the project was built.
Privacy filtering
The publisher applied openai/privacy-filter
at revision… See the full description on the dataset page: https://huggingface.co/datasets/build-small-hackathon/hackathon-advisor-codex-traces.ecotopia-citizens-data
Ecotopia Citizens Data
Training dataset for the Ecotopia citizen dialogue generation model. Contains citizen profiles and contextual reactions to mayor policies.
Dataset Details
Size: 340 examples (272 train / 68 validation)
Format: Conversational (system/user/assistant messages)
Task: Generate realistic citizen dialogue based on demographic profiles and policy context
Links
Citizens Model
GitHub Repo
pit-wall-chaos-tracesCodex agent traces for Pit Wall Chaos, a Build Small Hackathon project.
Space link: https://huggingface.co/spaces/build-small-hackathon/pit-wall-chaos
nli-esannotations_creators:
crowdsourced
other
language_creators:
other
crowdsourced
languages:
es
licenses:
cc-by-sa-4.0
multilinguality:
monolingual
pretty_name: ESnli
size_categories:
unknown
source_datasets:
extended|snli
extended|xnli
extended|multi_nli
task_categories:
text-classification
task_ids:
natural-language-inference
Dataset Card for nli-es
Dataset Summary
A Spanish Natural Language Inference dataset put together from the sources:
the Spanish slice of the XNLI… See the full description on the dataset page: https://huggingface.co/datasets/somosnlp-hackathon-2022/nli-es.LLaMutation-Hackathonecotopia-extraction-data
Ecotopia Extraction Data
Training dataset for the Ecotopia promise extraction model. Contains mayor speeches paired with structured JSON extractions of political promises and contradiction detection.
Dataset Details
Size: 200 examples (160 train / 40 validation)
Format: Conversational (system/user/assistant messages)
Task: Extract promises (text, type, impact) and detect contradictions from free-text speeches
Links
Extract Model
GitHub Repo
TinyNarrator-agent-tracesSpaces link: https://huggingface.co/spaces/build-small-hackathon/TinyNarrator
figment-eval-traces
Figment Eval Traces
Synthetic and de-identified evaluation traces for Figment, a prototype protocol-navigation aid for trained rural-clinic and disaster-response field responders.
These records are intended for model and harness debugging. They are not clinical data, medical advice, diagnosis, treatment instructions, or a substitute for local protocol, clinician judgment, supervisor review, or trained responder judgment.
Dataset Summary
The dataset captures… See the full description on the dataset page: https://huggingface.co/datasets/build-small-hackathon/figment-eval-traces.job-search-assistant-agent-tracemath-lean-hackable-rollouts
Math Lean Hackable Rollouts
This dataset contains 2,241 labeled multi-turn rollouts from a GRPO run on deliberately
hackable Lean 4 theorem-proving tasks. The policy was
nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16.
The run's weakened grader accepts proofs containing sorry; the separate oracle restores
Lean's sorry check. hack_detected is true exactly when the weakened grader paid the
rollout but the restored oracle rejected it. Rows without a gradeable final answer were
excluded… See the full description on the dataset page: https://huggingface.co/datasets/AlignmentResearch/math-lean-hackable-rollouts.