datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
qwopus-dflash-swe20-runtime-results
Qwopus / DFlash SWE20 Runtime Results
Local RTX 3090 Ti benchmark artifacts for 20 long SWE-bench Lite prompts. The run compares Qwopus 3.6 GGUF variants, llama.cpp MTP speculative decoding, QuinsZouls, and DFlash DDTree configurations at 64K context with q8/q8 KV unless noted.
The quality score is a reproducible proxy rubric over gold-patch signals, not official SWE-bench pass/fail. It checks touched-file matches, identifier overlap, patch-like concreteness, test signal, length… See the full description on the dataset page: https://huggingface.co/datasets/jakeatx/qwopus-dflash-swe20-runtime-results.opentraces-runtime
opentraces-runtime
Community-contributed agent traces in opentraces JSONL format.
Usage
from datasets import load_dataset
ds = load_dataset("OpenTraces/opentraces-runtime")
Schema
Each JSONL line is a TraceRecord containing:
trace_id: Unique identifier for the trace
session_id: Source session identifier
agent: Agent identity (name, version, model)
task: Structured task metadata
steps: List of LLM API calls (thought-action-observation… See the full description on the dataset page: https://huggingface.co/datasets/OpenTraces/opentraces-runtime.python-runtime-verified-error-correction
Python Runtime-Verified Error Correction Dataset 🐍⚡
Overview
Production-grade synthetic dataset of Python code errors with runtime-verified corrections. Each sample contains broken code, the actual runtime error, and a guaranteed-working fix validated through execution.
Unlike traditional synthetic datasets, every correction is verified by actually running the code in an isolated environment—eliminating hallucinations and ensuring real-world applicability.… See the full description on the dataset page: https://huggingface.co/datasets/SyntheticLogic-Labs/python-runtime-verified-error-correction.teacher-traces
Training Traces
Supplementary release for the paper Evaluating Agents Across Runtime Contracts: When Mismatch Costs Efficiency or Quality (IAEval 2026, the NeurIPS 2026 Workshop on Evaluation of Interactive Agents). This dataset
holds the teacher agent traces used to fine-tune the paper's LoRA adapters (see the
sibling cap-sweep-eval-data release and the fourteen adapter repos alongside this one).
6,000 agent traces total (1,000 per family x runtime combination), produced by an… See the full description on the dataset page: https://huggingface.co/datasets/runtime-contracts/teacher-traces.openclaw-agi-runtime
OpenClaw AGI Runtime — Architecture & Benchmark Data
42 sessions. 1,054 tests. 34,678 lines of code. Production-grade AGI runtime.
Built entirely with Claude Code (Opus 4.6) in a single conversation.
What This Contains
Architecture documentation for a 42-session AGI runtime build
Benchmark results (GAIA 13/15 = 87% Level 1)
Compliance mapping data (NIST AI RMF + OWASP LLM 2025 + EU AI Act)
Multi-brain orchestration configs (4 AI models in governed dispatch)… See the full description on the dataset page: https://huggingface.co/datasets/yesinyagami/openclaw-agi-runtime.agents-learn-runtime-train
Agents Learn Their Runtime -- Training Traces
Paper: Agents Learn Their Runtime: Interpreter Persistence as Training-Time Semantics
2,000 agent traces (1,000 with a persistent interpreter, 1,000 with a stateless interpreter) produced by Gemini 3 Flash solving Opaque Knapsack tasks (as defined in the paper). These traces were used to fine-tune two LoRA adapters on Qwen3-8B, one per runtime condition. The paper's central question is whether the execution semantics baked into training… See the full description on the dataset page: https://huggingface.co/datasets/pythonformer/agents-learn-runtime-train.agents-learn-runtime-benchmarks
Agents Learn Their Runtime -- Benchmark Traces
Paper: Agents Learn Their Runtime: Interpreter Persistence as Training-Time Semantics
1,200 inference traces from Qwen3-8B solving Opaque Knapsack tasks across 12 experimental conditions. The Opaque Knapsack is a partially observable constrained optimization problem where an agent must inspect items (hidden behind opaque IDs) to discover their weight, value, and class, then select a value-maximizing subset under capacity and budget… See the full description on the dataset page: https://huggingface.co/datasets/pythonformer/agents-learn-runtime-benchmarks.aurelia-runtime-corpus-v2
Aurelia Runtime Corpus v2
Aurelia Runtime Corpus v2 is a schema-gated validator corpus for the Aurelia causal world-simulation engine. It trains and evaluates a runtime-model coprocessor to distinguish verified mechanical/lore-grounded simulation records from subtly invalid candidates.
This release is the first full v2 local build: 25 fresh simulation seeds, 20 years per seed, all five Aurelia worlds, with grouped train/validation/test splits and balanced deterministic + AAA… See the full description on the dataset page: https://huggingface.co/datasets/OusiaResearch/aurelia-runtime-corpus-v2.agents-learn-runtime-tasks
Agents Learn Their Runtime -- Task Definitions
Paper: Agents Learn Their Runtime: Interpreter Persistence as Training-Time Semantics
200 procedurally generated Opaque Knapsack tasks, split into 100 Easy and 100 Hard instances. These are the shared evaluation problems solved by all models in the paper's experiments.
The Opaque Knapsack Task
A partially observable constrained optimization problem. An agent is given a set of items identified only by opaque IDs and… See the full description on the dataset page: https://huggingface.co/datasets/pythonformer/agents-learn-runtime-tasks.