datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
LOOPerSet
LOOPerSet: A Large-Scale Dataset for Data-Driven Polyhedral Optimization
Dataset at a Glance
LOOPerSet is a corpus of 28 million labeled compilation traces designed for machine learning research in compilers and systems. It maps synthetically generated loop nests and complex optimization sequences to ground-truth execution times measured on physical hardware. Transformation sequences were generated using a polyhedral compilation framework to ensure they… See the full description on the dataset page: https://huggingface.co/datasets/Mascinissa/LOOPerSet.LoopsBench
LoopsBench
This dataset repository hosts published LoopsBench task bundles. A LoopsBench task is a self-contained evaluation package for long-horizon terminal coding: it includes an agent-visible workspace snapshot, unit-level requirements, dependency graphs, Docker execution metadata, public verifier files, and reference gold patches used by maintainers and Oracle-style validation.
The files in this dataset are release artifacts mirrored from the latest LoopsBench GitHub… See the full description on the dataset page: https://huggingface.co/datasets/LoopsBench/LoopsBench.relay-ko-s2-scratch-g4-150h-R0-arelay-ko-s2-scratch-g4-53h-R0-arelay-ko-s2-scratch-g4-100h-R0-brelay-ko-s2-scratch-g4-53h-R0-brelay-ko-s2-scratch-g4-100h-R0-arelay-ko-s2-scratch-g4-150h-R0-bloopwan-opensora-pilot-v1
LoopWan Open-Sora-Plan pilot
Status: completed bounded curation. Counts: {"long_audit": 22, "train": 2000, "val": 128}.
Fixed 320x480, timestamp sampling at 16 FPS; train/validation crops are real
contiguous 10-second shots, audit crops 20 seconds. Sources are disjoint and
captions are matched to pinned official annotations. See DATASET_REPORT.md for
filter thresholds, caption limitations and full provenance.
Official dataset revision: ab77293def393e6938f11a7bfd12163decfb9620.… See the full description on the dataset page: https://huggingface.co/datasets/Nicholas0228/loopwan-opensora-pilot-v1.LoopTF-Sudokucircle-packing-insight-loop
Circle-Packing Insight-Exploration Loop
Artifacts from an iterative GPT solver <-> proposer insight-exploration loop on the
21-circles-in-a-perimeter-4-rectangle packing problem (AlphaEvolve SOTA sum-of-radii
= 2.3658321334167627). Each round, 16 solvers propose a program + written explanation;
every program is scored; a proposer then mines all 16 attempts into an evolving insight
document that conditions the next round. Run: 16 solvers x 8 rounds.
Subsets… See the full description on the dataset page: https://huggingface.co/datasets/ars22/circle-packing-insight-loop.LoopTool-23k
Overview
LoopTool is a fully automatic, model-aware iterative framework that tightly couples data generation and model training for tool-augmented LLM learning
The LoopTool-2w is released as part of Closing the Data–Training Loop for Robust LLM Tool Calls
The dataset comprises 23,040 tool-call samples, involving 20,813 APIs. In each sample, the instruction contains the corresponding set of available tools for that sample; the input corresponds to the dialogue history of the… See the full description on the dataset page: https://huggingface.co/datasets/zhangkangning/LoopTool-23k.behavioral-loops
Behavioral Loops
1,140 behavioral patterns across 279 categories, each structured as given/when/then/result logic with taxonomy classification, veracity scores, and intervention strategies.
Quick Start
from datasets import load_dataset
ds = load_dataset("buley/behavioral-loops")
print(ds["train"][0])
Structure
Field
Description
given
Initial condition or context
when
Trigger event
then
Resulting behavior
result
Long-term outcome
origin… See the full description on the dataset page: https://huggingface.co/datasets/buley/behavioral-loops.relay-ko-s2-scratch-53h-R1-brelay-ko-s2-scratch-53h-R1-ahedgehog-loop-control-r4
hedgehog-loop-control-r4
Hedgehog — loop-control round 4 (termination/repetition fixes).
Contents
train.jsonl (2944 rows)
validation.jsonl (438 rows)
Format
JSON Lines (.jsonl), one example per line.
Provenance
Original content for the Hedgehog extraction model (Michael Anthony Falabella).
looped_llama_v3_datasetLoopHard
LoopHard
LoopHard is a held-out benchmark of 285 English enumeration prompts designed to
measure futile reasoning loops in thinking models. The primary metric is
judged loops: the percentage of traces that end while still repeating the
same recall, verification, or enumeration cycle without adding or settling
information.
Long or systematic reasoning is not automatically a loop. A verification pass
that converges is not a loop, and a trace that breaks out of a loop on its own
is… See the full description on the dataset page: https://huggingface.co/datasets/N8Programs/LoopHard.relay-ko-s2-scratch-53h-R0-arelay-ko-s2-scratch-53h-R0-bloop-qwen-v8-sft
loop-qwen-v8 SFT dataset (Gemini insulin-control distillation)
8,222 chat-format examples used to SFT loop-qwen-v8 (Qwen3-4B insulin
controller distilled from Gemini-3-flash-preview). Each example is a closed-loop
dosing decision.
Format (JSONL, one chat per line)
system: controller spec (IOB-aware, Chain-of-Draft reason-before-act)
user: patient metadata (age, weight, TDD, CF, IC, basal) + 6h history of CGM / insulin / carbs, as JSON
assistant: JSON… See the full description on the dataset page: https://huggingface.co/datasets/jxx123/loop-qwen-v8-sft.humanoid-cognitive-interaction-loops
Humanoid Cognitive Interaction Loops
Closed-loop cognitive interaction cycles for humanoid agents.
ptdbench-data-format-task-agent-loop-022-llama-dapo-math-dataset
PTDBench dataset snapshot: task_agent_loop_022-llama-dapo-math
This repository stores the immutable runtime dataset snapshot for one
materialized PTDBench task. It intentionally excludes model weights and
training checkpoints.
PTDBench family: data_format
Source evaluation metric: val-core/math_dapo/reward/mean@1
Provenance: Processed from BytedTsinghua-SIA/DAPO-Math-17k; task-specific bytes are pinned.
License: Apache-2.0
The artifact manifest records every hydrated runtime… See the full description on the dataset page: https://huggingface.co/datasets/LIF1014/ptdbench-data-format-task-agent-loop-022-llama-dapo-math-dataset.audio-loop
Audio Loop — Long-Audio Reasoning Benchmarks
Two evaluation sets for long-form audio reasoning, packaged as one dataset
with two configs:
Config
Task
Audio source
Items
aae_tts
Argument structure & contradiction
TTS-synthesized argumentative essays
20
iq2_qa
Multi-hop QA over debates
Recorded live debates
13
Load annotations
from datasets import load_dataset
aae = load_dataset("audioloop/audio-loop", "aae_tts", split="test")
iq2 =… See the full description on the dataset page: https://huggingface.co/datasets/audioloop/audio-loop.han-human-assistance-feedback-loop-v1
Human Assistance Feedback Loop Dataset
This dataset captures simple human feedback
after a humanoid robot completes an assistance task.
The goal is to study how feedback can be used
to refine future robot behavior.
Research Motivation
Robots should improve not only from success,
but also from subtle human reactions.
Use Cases
Feedback-driven learning
Human-in-the-loop robotics
Assistive behavior refinement
Part of
Humanoid Network (HAN)… See the full description on the dataset page: https://huggingface.co/datasets/ariefansclub/han-human-assistance-feedback-loop-v1.do_while_loop_cpp_to_pygoodsmash-fee-loop-stateloop-distill-data
loop-distill-data — Gemini→Qwen insulin-control distillation SFT (simulator)
Chat-format SFT (system / user-state / assistant action+reasoning) distilling a
gemini-3-flash-preview insulin controller. All states here are from the
simglucose simulator (virtual patients) — no real-patient data. IOB is provided
in the prompt; actions = basal/bolus chunks + a concise clinical rationale.
file
rows
provenance
distill_sft_v15.jsonl
10,182
simglucose rollouts + Gemini labels… See the full description on the dataset page: https://huggingface.co/datasets/jxx123/loop-distill-data.simpleRoute6open_loop_pano_filter_v2_panorama