LLM Chain-of-Thought Contains Surprising Emotionally Expressive Language—and It May Be Functional
xuanalogue · x · 2026-09-30
Drucker and kmahowald published a piece on expressives in LLM agents' chains of thought.
- They document surprising, emotionally expressive language appearing in AI "thinking" during math and hacking tasks
- The authors argue this expressive language is likely functional rather than noise, with implications for understanding what CoT actually does
More from Models
- Missed OpenAI DevDay? A 'Keynote in 5' recap gets you up to speed — craigsdennis · 2026-09-30
- Dev team cut $3k token budget with GPT subs, and the workaround is already slipping — natesiggard · 2026-09-30
- Free ChatGPT users can't see how far AI has come, hurting conversion, argues dev — Angaisb_ · 2026-09-30
- GPT-6.1 Sol beats Pokemon Red's first gym in 244 turns, a PokeBench record for $2.60 — VibeCodyH · 2026-09-30
- Gemini voice bug: chosen accent switches when making calls via 'Hey Google' — Rhath223 · 2026-09-30
- Sentdex: GLM 5.3 Puts a Frontier Model on Your Home Machine — Stop Paying Companies to Lecture You — Sentdex · 2026-09-30