Long-context models still waste tokens by copying the prompt into reasoning
eyishazyer · x · 2026-07-23
A post about a paper argues that long-context reasoning still has a hidden failure mode: even strong models copy large chunks of the prompt into their own chain-of-thought, wasting tokens and hurting accuracy.
- The assumption that long context solved retrieval is challenged.
- Copying gets worse as inputs get longer.
- The result quietly eats token budget and degrades performance.
- The proposed fix is not just more context, but smarter context handling.
Related event: Study Reveals Long-Context Models Waste Tokens Copying Text(2 posts)→
More from Research
- Paper evaluates speech translation by whether listeners can answer key questions — EhudReiter · 2026-07-23
- Alignment paper warns static preferences can lock in values and stall societies — edelwax · 2026-07-23
- AgenC now defaults to one agent after multi-agent systems fell 39%–70% behind — tetsuoai · 2026-07-23
- LLMs may help most in the manual refinement phase of decompilation — OwariDa · 2026-07-23
- RLSS 2026 in Milan turns Bellman backups into a coffee-fueled group sport — misovalko · 2026-07-23
- RLSS 2026 Milan masterclass on world models and RL spans 23 slides — misovalko · 2026-07-23