Daily papers of 28 Oct 2026

Daily Papers

Here is the selection of papers for today (28 Oct):

• Multiscale Policy Optimization for Long-Horizon LLM Agents (373 ▲)
• ChronoReplay: One Answer Interface Is Not Enough for Streaming Spatial Intelligence (345 ▲)
• The Trinity:Capability-Driven Recursive Self-Improvement for Multimodal Agents (312 ▲)
• VibeVideoBench: Evaluating Multimodal Agents on Code-Driven Video Generation (172 ▲)
• PDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations (136 ▲)
• Memory Tax: Dissecting the Net Value of Agent Memory (124 ▲)
• Hy-Nomas: Towards Free-form Interactive World Generation (105 ▲)
• Can Test-Time Memory Evolution Outperform Test-Time Training for Scientific Discovery? (101 ▲)
• Can Discrete Tokens Serve as Semantically Lossless Representations of Visual Signals? (82 ▲)
• HABIT-Bench: Evaluating Habit Induction from Longitudinal Weak Evidence in Agent Memory (79 ▲)
• Rollback: Experience Backtracking in World Models for Policy Improvement (51 ▲)
• CodeCap-Bench: Can Code Serve as a Good Video Caption? (49 ▲)
• And 39 more papers

NEW - I'm opening Daily Papers! If you're a researcher you can now submit your peers' papers to be featured in Daily Papers directly. Use this page to do so: huggingface.co/papers/submit.

Onwards and upwards,
AK and the research community