Report finds LLMs surprisingly good at reading and writing neural network weights
omarsar0 · x · 2026-11-03
A must-read report highlighted by Elvis Omarsar finds that LLMs are surprisingly good at reading and writing neural network weights directly. He argues this is huge for recursive self-improvement (RSI): current RSI pipelines focus on writing code to launch training jobs, but weight manipulation could offer a more native path for models to recursively improve themselves.
More from Research
- LLM CoTs are often post-hoc rationalization; CRoS reward boosts faithfulness by 13.8% — mohitban47 · 2026-11-03
- Why powerful AIs could cooperate in one-shot prisoner's dilemmas: open-source code proofs — tszzl · 2026-11-03
- Paperclip Launches Datasets: Agents Build Data Tables From Papers With Source Citations — james_y_zou · 2026-11-03
- NVIDIA's Dory1-4B: A Looped LLM With 8 Input, 20 Recurrent and 8 Output Layers — scaling01 · 2026-11-03
- Nvidia ships Dory1-4B, a looped LLM that reuses 20 layers to rival much larger models — scaling01 · 2026-11-03
- Google and Cornell Use Sleep-Inspired MoE Distillation to Fix Catastrophic Forgetting — jmugan · 2026-11-03