How Language Models Fail: Token-Level Signatures of Committed and Persistent Reasoning Failures
For researchers and practitioners using LLMs, this provides a framework to detect and adapt to reasoning failures, improving reliability.
The paper identifies two distinct failure modes in language model reasoning—committed failure and persistent uncertainty—using token-level uncertainty signals, and shows these signatures generalize across 23 model-dataset configurations with predictions holding in 20 of 23 cases.
Failures in language model reasoning emerge through distinct processes that leave identifiable signatures in the reasoning trace. We characterize these failures using token-level uncertainty signals, finding they arise through two empirically distinguishable processes. The first is committed failure, in which a model locks onto an incorrect reasoning path early in its trace. A central diagnostic signature is the commitment point, beyond which considering additional tokens hurt rather than help failure detection. In the second, persistent uncertainty, uncertainty instead accumulates throughout, and the full trace is needed to best distinguish failing from successful completions. These signatures reproduce across 23 model-dataset configurations, with the framework's falsifiable predictions holding in 20 of 23 cases, well above chance across both failure modes. Finally, we demonstrate our failure mode framework has direct implications for self-consistency, identifying when uncertainty signals complement it and when it can be selectively skipped. These results offer a foundation for understanding when LLM reasoning failures become detectable and for adapting detection strategies accordingly.