ClueFlow: Learning Long-Horizon Search through Attributed Research-State Transitions
Attributes search progress to research-state transitions instead of turn-level signals.
AI Storyline
3 items · 2 sources · 3 days
Operational story trace
Learning Through
Follow in this browser to see new updates on your Live feed.
Latest change
A multimodal RL paper on video editing asks when fine-grained credit assignment helps agent training and when it fails to improve task performance.
Earlier contextThe story so far
A run of arXiv papers on how agents learn from their own trajectories. ClueFlow credits search steps by research-state transitions, and a memory-skill paper tackles lifelong learning under a bounded context window.
editor-curated · source-linked
Arc
Oct 19Oct 26 · now
SEARCH · Oct 19ClueFlow credits long-horizon search by research-state transitions1 source · show source ▾
MEMORY · Oct 23Bidirectional memory-skill learning for lifelong agents1 source · show source ▾
Lifelong Context Adaptation through Bidirectional Memory-Skill Learning
Couples memory with learned skills so decisions fit a bounded context.
NOW · Oct 26Fine-grained credit assignment helps and hurts in multimodal agentic RL1 source · show source ▾
When Fine-Grained Credit Assignment Helps and Hurts: Multimodal Agentic Reinforcement Learning through Video Editing
Identifies conditions where fine-grained credit assignment fails to improve task performance.
Day 1 Monday, Oct 19, 2026
ClueFlow: Learning Long-Horizon Search through Attributed Research-State Transitions
Attributes search progress to research-state transitions instead of turn-level signals.
Day 2 Friday, Oct 23, 2026
Lifelong Context Adaptation through Bidirectional Memory-Skill Learning
Couples memory with learned skills so decisions fit a bounded context.
Day 3 Monday, Oct 26, 2026
When Fine-Grained Credit Assignment Helps and Hurts: Multimodal Agentic Reinforcement Learning through Video Editing
Identifies conditions where fine-grained credit assignment fails to improve task performance.
What to watch — open questions
- Do ClueFlow-style state-transition rewards hold up beyond the search benchmarks in the paper?
- Which task properties predict that fine-grained credit assignment will hurt rather than help?
How this thread was built
editor wrote the arc · 3 beats
Storylines are threaded mechanically from the feed: stories that share a distinctive anchor across multiple days and sources. Each item links to its original source. The evidence trace, current state, and open questions are written by the editor routine and refreshed whenever a new beat lands.