Researcher blames swarm RL mindset and token-efficiency pressure for unreliable models
tokenbender · x · 2026-11-18
Replying to tszzl (who disliked the well-loved 5.6 Sol but loves the initially-hated 6.2 Astra), tokenbender argues the current training paradigm — swarm RL mindset plus token-efficiency pressure — keeps reliability shitty while piling on decision and comprehension fatigue, unless the failure to improve ML research capabilities is chalked up to strategic reasons.
More from Models
- Google Ships Gemini 4 Neon, Positioned Between Carbon and Hydrogen for Fast Agentic Work — jeremyphoward · 2026-11-22
- Microsoft's USB4 RDMA lets mini PCs pool memory to run Qwen3-235B — JFPuget · 2026-11-22
- Google AI Mode now enforces usage limits, users surprised by rate-limit popup — elvano124 · 2026-11-22
- Sol 6.2 now refuses reverse engineering that Sol 6.1 happily did for a month — cardak98 · 2026-11-22
- Anthropic's record ~$81B IPO, GPT-6.2 pricing, Qwen4-Max and Copilot Autopilot: week in review — ksraj1001 · 2026-11-22
- Reddit rant: Frontier AI labs are training models to imitate lazy engineers — RossPeili · 2026-11-22