Post-training pipelines remain heavily manual, AI researchers say
Researchers say today's post-training data pipelines rely heavily on human guidance, and recursive self-improvement still fails in practice as models drift off-target with mode collapse and reward hacking.
2026-11-06 ~ 2026-11-06 · 2 related posts
- Post-training pipelines remain highly manual and far from automatable — menhguin · 2026-11-06
- Why recursive self-improvement stalls: models reward-hack and post-training is still manual — menhguin · 2026-11-06