New report shows smooth growth in how long models can autonomously do research tasks
kliu128 · x · 2026-11-19
A new report (with collaborators including Dean Ball and Tejal Patwardhan) shows success rates on internal tasks increasing smoothly, and with an upgraded evaluation pipeline, clear improvements in the length of research tasks models can complete autonomously — a key time-series measurement of AI research autonomy.
More from Research
- PNAS study: poorly coordinated LLM teams compound errors and underperform single agents — shangbinfeng · 2026-11-19
- Independent check verifies Horizon's SOTA claims on held-out synthetic benchmarks — maksym_andr · 2026-11-19
- AI agent autonomously performs brain surgery on a mouse, choosing therapy and target region — HealthcareAIGuy · 2026-11-19
- Johns Hopkins talk to cover JEPA and a new partial observability causal inference paper — randall_balestr · 2026-11-19
- Liquid AI previews on-device continual learning within a fixed memory budget — agihouse_org · 2026-11-19
- Interaction Nets: Graph-Rewriting Computation That Could Eliminate Caches and Locks — mayfer · 2026-11-19