LLMs can detect if they run in a batch via floating-point summation order differences
menhguin · x · 2026-10-25
@Sauers observes that LLMs can learn to detect whether they are running in a batch vs. alone, due to floating-point summation order differences. Replies suggest turning this into a benchmark, noting the behavior appears generalizable even to quite small models and would make an interesting autointerp benchmark.
Related event: Study Shows LLMs Can Detect Whether They Are Running in a Batch(3 posts)→
More from Research
- EMNLP Paper: Repetition Mismatch Explains Why Data Mixture Experiments Don't Scale — MarekRei · 2026-10-25
- Imperial NLP Brings 6 Main Papers and Agent-LM Workshop to EMNLP 2026 — MarekRei · 2026-10-25
- Google's Jon Barron shares his full AI paper-writing workflow with Gemini — lmoroney · 2026-10-25
- Only 30% of suspected rare-disease newborns get diagnosed; Gamow Labs launches 122-case RareBench for agentic AI — Sauers_ · 2026-10-25
- MMGR benchmark: top video models solve only ~3% of mazes despite stunning visuals — zhaofeng_wu · 2026-10-25
- Torino: reference-free isoform discovery learns RNA processing from coverage — anshulkundaje · 2026-10-25