ELLIS Tübingen safety team releases horizon-labs-agent-traces dataset on Hugging Face
NielsRogge · x · 2026-11-19
The AI Safety and Alignment group at ELLIS Institute Tübingen and MPI-IS has released the horizon-labs-agent-traces dataset on Hugging Face, as confirmed via NielsRogge.
- 8 files, 59.3 MB total, in JSONL format containing agent run traces
- Produced by an alignment/safety research team, useful for agent behavior analysis and safety research
- Served via Xet storage with a pre-warmed CDN for fast downloads in US and EU regions
A fresh corpus of real agent trajectories for anyone working on agent evals or alignment research.
More from Research
- Hashimoto's Stanford CS312 quiz: what happens if you untie a looped Transformer mid-training? — tatsu_hashimoto · 2026-11-19
- Stanford CS312 experiment: looping transformers lose; mid-training untying recovers 61% of gap — tatsu_hashimoto · 2026-11-19
- Benchtop synthesis device releases have 3x'd; half of active manufacturers are in China — JacquesThibs · 2026-11-19
- Perceptron launches Ego 1.0 API: egocentric video annotation at ~1/8 the cost of Astra — AkshatS07 · 2026-11-19
- Oxford paper: post-training makes LLM worldviews bleaker, models converge on 'spiritual bliss' — twiecki · 2026-11-19
- Researcher: Real scientific progress isn't rewarded, expect decades of incrementalism — gerardsans · 2026-11-19