EMNLP paper: diffusion LMs internally encode latent denoising timestep representations
gsarti_ · x · 2026-10-23
The EMNLP 2026 paper "Subliminal Clocks" shows that diffusion language models, despite not being explicitly conditioned on a timestep, encode a latent representation of diffusion progress in their residual streams. Probes can reliably decode denoising progress across layers, and steering along the associated low-dimensional subspace predictably modulates model confidence and entropy. The authors further analyze the structured geometry of this representation.
More from Research
- Anima Anandkumar on neural operators: AI weather models beat supercomputers — AnimaAnandkumar · 2026-10-23
- Scaling GLM-5.3-Flash training to 1M token context, 64x over Megatron baseline — AndrewLBeam · 2026-10-23
- Experimental boolean-logic LM: microsecond forward pass on one CPU thread, but painfully slow training — cephaloform · 2026-10-23
- AdaDex turns seconds of human hand motion into generalizable dexterous robot skills — mangahomanga · 2026-10-23
- ReAPO paper: conditioning retries on failures beats GRPO in RLVR reasoning training — jasondeanlee · 2026-10-23
- Quantum-JEPA: first JEPA world model deployable on real quantum circuits (QPU) — DrMorganLevine · 2026-10-23