Must-read papers of the week: continual learning, world models, agent serving
TheTuringPost · x · 2026-11-03
The Turing Post's weekly paper roundup, heavy on agents and world models:
- HaMo: harness-model collaboration for long-horizon continual learning (Huzhou Normal Univ./Alibaba et al.)
- Rollback (experience backtracking in world models), CausalWAM (causes of actions), MiRL (error-targeted RL for agent misbehavior), WebAnything (wild web as agent environments)
- Controlled study of visual memory in long-video generation; can LLMs uncover hidden physical laws via free exploration
- TAO: tool-aware joint scheduling for agentic LLM serving; REA: routed embodied agent runtime; ThinkWorld: world imagination for chain-of-dynamics reasoning; DecoyCount visual-reasoning benchmark.
More from Research
- Direct Weight Editing via Codex+Astra Lifts Audio Classifier Accuracy From 84.3% to 88.75% Without Retraining — dansitu · 2026-11-03
- How UMI turns handheld grippers and SLAM into robot training data, in 5 steps — gabriberton · 2026-11-03
- TSFLab: A Self-Improving Agent-Agnostic Workbench for Time Series Forecasting — D2I-Group · 2026-11-03
- YC's Momentic launches QA-Bench: its AI QA agent finds 2.7x more reproducible bugs than Codex — ycombinator · 2026-11-03
- EDU-Jev 4B education decision model hits 67.6%, beating top commercial API by 9.4 points — kalyan_kpl · 2026-11-03
- ReAct to Production Agents: The 7 Layers Between Model and Result — blaizedsouza · 2026-11-03