Why RL generalizes to reasoning but not literary writing, per AI researchers
phl43 · x · 2026-09-22
phl43 argues that even the latest models, while imperfect on deep conceptual issues, do far better there than on literary work — likely thanks to generalization from RL on easier goals. Doing the same for literature seems genuinely harder and costlier. The practical workaround discussed: rubrics, examples, and step-by-step bootstrapping, which works but is slow going.
More from Models
- Average users aren't throwing frontier models at open math problems, dev observes — felpix_ · 2026-09-22
- Zero-day hits Meta's Muse for Mac: local process can steal prompts, auth tokens and file access — MicahBerkley · 2026-09-22
- Grok 4.7 spotted in user chatter as users hope for usage reset — BWay124 · 2026-09-22
- Amassing a PhD team is exactly what OpenAI did, dev notes in AI research debate — felpix_ · 2026-09-22
- Dev argues prompting alone can't get AI to solve natural science problems — felpix_ · 2026-09-22
- Dev claims benchmarks are 'absolutely meaningless' — models only differ by vibe — gnukeith · 2026-09-22