Extended reasoning suppresses scalar implicatures: models entertain then cancel not-all readings
Zhizhuo Kou · hf · 2026-10-22
HKUST researchers study how extended thinking affects everyday pragmatic inferences like the scalar implicature that "some students passed" implies not all did.
- Keeping weights fixed and varying only thinking length, they record which reading models choose
- Extended reasoning suppresses everyday scalar implicatures: models initially consider enriched interpretations (not all, not both) during thinking, then cancel them before answering literally
- A second tool probes the model's own reasoning at successive cut points, locating where the reading flips — early-stage answers can still recover the pragmatic inference
More from Research
- 36 preprints in 3 months: BootLoops AI harness drives cross-field science — DeryaTR_ · 2026-10-22
- Meta and MSU paper: LLM-built reasoning graphs are unreliable and their errors leak into confidence scores — rohanpaul_ai · 2026-10-22
- New paper: can first-price auctions efficiently allocate queries among strategic AI agents? — _onionesque · 2026-10-22
- RUC's LoopRec Brings Item-Level Adaptive Recursion with Early-Exit Gates to Sequential Recommendation — Zhanshuo Cao · 2026-10-22
- Anthropic's Agent4Sci: Frontier Agents Beat SOTA on 20 Scientific ML Benchmarks, but Only via Engineering Tricks — Anthropic · 2026-10-22
- NUS Paper Proposes PAPAV: A Unified Capability Framework from Computer-Use to Robot-Use Agents — showlab · 2026-10-22