NeurIPS 2026 journal track: 88 papers, ASU's "Beyond Semantics" among TMLR picks
karthikv792 · x · 2026-10-18
NeurIPS 2026 announced its Journal Track accepts: 64 TMLR, 16 JMLR, and 8 Annals of Statistics papers, to be presented in Sydney, Atlanta, and Paris this December. Among them is ASU's "Beyond Semantics," showing transformers trained on corrupted or semantics-free reasoning traces perform comparably to those trained on correct traces, generalize better OOD, and GRPO boosts accuracy without improving trace validity — challenging the assumption that CoT tokens reflect real reasoning.
More from Research
- Stanford's RT-J: 86M relational foundation model trained on 639 web-crawled databases — christopher · 2026-10-18
- Optimus launches: an autonomous RL agent harness that runs experiments itself — _AustinCalvert_ · 2026-10-18
- Fields Medalist Timothy Gowers: one-shotting a proof with ChatGPT means we don't understand it — haider1 · 2026-10-18
- NVIDIA to present CANTO: tokenizing CAD geometry for aerodynamics prediction — JeanKossaifi · 2026-10-18
- Papers with Code adds Auto-Research category, 166 papers already tracked — NielsRogge · 2026-10-18
- ModelTrace: open-source tool fingerprints LLMs with 3 long-integer challenges — scaling01 · 2026-10-18