SAS Makes Schedulers Self-Aware
furongh · x · 2026-07-06
The core idea is to give the scheduler "self-awareness": SAS leverages the frozen model's own path likelihood to reward a specific unmasking sequence—essentially asking, "How well can I explain the target along this path?". This allows the model to learn a decoding order that naturally aligns with its own predictive strengths.
Related event: ICML 2026 Paper SAS: Optimizing Thought Scheduling in Diffusion LMs(15 posts)→
More from Research
- LightMTP: lightweight latent multi-token prediction boosts LLM planning in pretraining — gdm3000 · 2026-10-27
- EMNLP study: AI literary translation reads fine but varies more, even within one book — tuvllms · 2026-10-27
- Lab monkeys at $30-80K each: a 4-year breeding cycle is the real preclinical drug bottleneck — zebird0 · 2026-10-27
- OpenAI Foundation funds UNC tumor antigen project, hiring senior experimental scientist — iskander · 2026-10-27
- Steering attention heads toward honesty: up moral patienthood, down psychopathy — edelwax · 2026-10-27
- Are transcoders overkill? New paper argues plain neurons suffice for circuit analysis — burny_tech · 2026-10-27