MaLoW internalizes streaming video history into LoRA weights, beating prior memory methods
uw-madison · hf · 2026-10-16
Researchers at UW–Madison and Peking University propose MaLoW, a streaming memory framework that internalizes video history into lightweight parameter updates instead of compressing or caching observations.
- A recurrent HyperNetwork incrementally consolidates incoming video frames into fixed-size LoRA weights — "history as weights" — without revisiting the full stream
- Introduces MaLoW-Instruct: 381K QA pairs over 56K videos
- Outperforms prior memory approaches by 3.9%–11.8% on OVO-Bench's backward split with modest latency overhead; also generalizes to VLA models, improving Simpler WidowX action success by 9.3%
More from Research
- Mosaic: multi-session agentic RL with hierarchical advantage estimation — LARK-Lab · 2026-10-16
- Meituan LongCat's WorldCast benchmark: GPT-Image-2.5 tops at 72.8, open-source lags — meituan-longcat · 2026-10-16
- Recursive Environment Scaling tanks six models' success rates by 57–77 points — Lingxiao Jing · 2026-10-16
- Harvard's MPT: decentralized agent swarms learn self-orchestration via bidding — Harvard · 2026-10-16
- MIT Biology and Schwarzman College hiring tenure-track professor in computational biology with generative AI focus — anshulkundaje · 2026-10-16
- Developer live-runs continual pretraining of a Luganda LLM on Colab, 41% done with loss dropping fast — weskambale · 2026-10-16