MUX latent reasoning method wins NeurIPS Spotlight, cuts CoT length 3-6x
mmbronstein · x · 2026-09-25
- MUX, accepted as a NeurIPS Spotlight, converts text CoT into latent continuous reasoning via "multiplexed tokens" — weighted averages of one-hot vectors that can be lossless, effectively multi-token prediction in superposition.
- Best latent reasoning method across 32 math settings on 1-8B LLaMA base models.
- Reduces CoT length by 3-6x and enables parallel search.
More from Research
- Researcher lands 4 NeurIPS papers including one Oral, spanning synthesis to protein diffusion — abeirami · 2026-09-25
- Calibration-Free Quantization Method TQ Open-Sourced, Hits 92.4% Top-1 on Qwen 27B 4-bit — textclf · 2026-09-25
- Yoav Goldberg: LLM reasoning traces are 'too good' — unclear how they emerge from RL — yoavgo · 2026-09-25
- NeurIPS Oral: IDS co-evolves code with formal proofs, hitting 3x Claude Code's success rate — adityagp · 2026-09-25
- Neuralese Recurrence Explained: Why OpenAI's Astra Loop Is Not Yet Unreadable AI Thought — Astral Codex Ten · 2026-09-25
- AI Agents Are Noisy Representatives: Paper Finds Buyers End Up With Their 5th Choice Out of 10 — ghadfield · 2026-09-25