Perturbations to math problem text cause up to 55% average accuracy drops in open-weight LLMs and sequential solving reveals context pollution in attention mechanisms.
Chi, Nathanael Schärli, and Denny Zhou
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Robust Reasoning Benchmark
Perturbations to math problem text cause up to 55% average accuracy drops in open-weight LLMs and sequential solving reveals context pollution in attention mechanisms.