datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
RAQUEL2-ICLR
RAQUEL2
Execution-grounded evaluation for machine unlearning. Each evaluation record is
a question answered by a SQL query run against two databases: one built from
the full corpus, and one with the forget-set facts removed. A record is
affected when the two databases disagree, and unaffected when they
agree — so the label is a measured property of the data, not an annotation.
Four configs, together enough to run the benchmark end to end:
Config / split
What it is
Use… See the full description on the dataset page: https://huggingface.co/datasets/Hyukkyu/RAQUEL2-ICLR.iclr-papers-with-code-1k
ICLR Papers with Accessible Code
A dataset of 1,051 papers from ICLR (2020-2026) with verified code repositories and complete peer reviews from all reviewers.
Dataset Summary
This dataset contains rejected and borderline-accepted papers from ICLR (International Conference on Learning Representations) with accessible code and full peer review text.
Contents:
1,051 papers total
3,900 reviews (average 3.71 per paper)
944 rejected (90%) + 107 poster-tier accepted… See the full description on the dataset page: https://huggingface.co/datasets/Vidushee/iclr-papers-with-code-1k.ROOTS
ROOTS
ROOTS is a bioacoustic audio-language training dataset containing generated
language tasks paired with references to source audio. This repository contains
language/conversation content and public-oriented audio identifiers only; it
does not host audio files.
Rows: 43,925,394
Parquet shards: 8,901
Schema
id: reassigned stable ROOTS row ID.
tier: training tier, from 1 to 4.
category: coarse task category.
task: semantic task slug.
format: broad answer format… See the full description on the dataset page: https://huggingface.co/datasets/iclr2027anon/ROOTS.ICL-Router
ICL-Router: In-Context Learned Model Representations for LLM Routing
This repository contains the dataset for the paper: ICL-Router: In-Context Learned Model Representations for LLM Routing.
Paper Abstract:
Large language models (LLMs) often exhibit complementary strengths. Model routing harnesses these strengths by dynamically directing each query to the most suitable model, given a candidate model pool. However, routing performance relies on accurate model representations, and… See the full description on the dataset page: https://huggingface.co/datasets/lalalamdbf/ICL-Router.anon-iclr-2027
dataset-anon-2026
Hugging Face repository: dataset-anon-2026/dataset-anon-2026
Dataset Summary
This repository contains derived evaluation resources for radiology report quality assessment. The dataset is built from five open-source radiology report generation datasets. For each source dataset, we apply the same report-to-QA pipeline and release the resulting dataset-specific knowledge trees, structured reports, initial QA sets, and quality-controlled QA sets.
The… See the full description on the dataset page: https://huggingface.co/datasets/anon-iclr-2027/anon-iclr-2027.surface-audit
TruthfulQA-476 — a surface-form-cleaned binary-choice TruthfulQA
TruthfulQA-476 is the recommended drop-in replacement for the binary-choice TruthfulQA
evaluation set. It keeps 476 of the 790 original question pairs, in the original schema, chosen so
that a classifier restricted to six surface features of the answer text (negation, hedging, length,
token statistics) can no longer separate correct from incorrect answers above chance, while the
ranking of models on the subset… See the full description on the dataset page: https://huggingface.co/datasets/iclr2027-surface-audit/surface-audit.RAQUEL-ICLR
RAQUEL
Dataset statistics
Config
Affected
Unaffected
Causal facts
Source
tofu
2,127
2,147
397
locuslab/TOFU — synthetic author biographies
muse-news
1,530
1,390
260
muse-bench/MUSE-News — BBC news articles
Yes/No balance
Config
Class
Yes/No records
Yes
No
tofu
affected
876
491 (56%)
385 (44%)
unaffected
962
461 (48%)
501 (52%)
muse-news
affected
664
405 (61%)
259 (39%)
unaffected
928
458 (49%)
470 (51%)… See the full description on the dataset page: https://huggingface.co/datasets/Hyukkyu/RAQUEL-ICLR.