Daily Papers of 2026-09-08

1. Unlocking Lossless Speedups in LLMs via Discrete Diffusion 150 upvotes, #1 of 2026-09-08
2. FlowBalance: Verifier-Grounded Self-Improvement from On-Policy Reasoning Experience 98 upvotes, #2 of 2026-09-08
3. ENEAS: Embedding-guided Neural Ensemble for Adaptive Segmentation 47 upvotes, #3 of 2026-09-08
4. Causal Foundation Models 31 upvotes, #4 of 2026-09-08
5. Verify Before You Distill: Prompt-Level Teacher Gating for On-Policy Distillation 20 upvotes, #5 of 2026-09-08
6. One Symptom, Three Levers: A Critical Review of On-Policy Self-Distillation 17 upvotes, #6 of 2026-09-08
7. EmbodiedSkills: A Unified Framework for Orchestrating, Training, and Deploying VLA Agents 16 upvotes, #7 of 2026-09-08
8. Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal 12 upvotes, #8 of 2026-09-08
9. SimpleMemVLA: A Simple but Effective Native-Video Memory for Vision-Language-Action Models 11 upvotes, #9 of 2026-09-08
10. Unifying Conformal Language Tasks with In-Context Ensembles 10 upvotes, #10 of 2026-09-08
11. What Else Needs Fixing? Exploring Cost-Effective Test-Time Compute for Revision Propagation in Artifacts Generated Through Conversation 6 upvotes, #11 of 2026-09-08
12. Privacy Failure in Split-LLM Training, The Returned Gradient Nullifies the Decoys 5 upvotes, #12 of 2026-09-08

Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.