LLMs Can See the Smoke but not the Fire: Evaluating Abductive Reasoning with Elenchos

2026-07-14

AI Summary

The capacity of large language models (LLMs) to perform abductive reasoning—specifically, inferring underlying rule changes from observed behavioral discrepancies—remains poorly understood. This work proposes Elenchos, an evaluation framework that formalizes abductive reasoning as a structural inverse problem: given behavioral traces from both original and perturbed versions of a formal system (e.g., λ-calculus), models must determine whether a rule modification occurred and precisely localize the change. Elenchos introduces a generative evaluation paradigm grounded in behavioral divergence between formal systems. Experimental results reveal that while mainstream LLMs can detect that a system has been altered, they struggle to accurately attribute the specific rule modifications, particularly when multiple interacting changes are present. Moreover, increasing computational budgets for reasoning yields only marginal performance gains.

Abstract

Large language models (LLMs) excel at pattern recognition and text generation, but their capacity for abductive inference - inferring latent hypotheses that explain observed behavior - remains poorly understood. Here, we introduce Elenchos (named after the Socratic method of cross-examination), a generative evaluation framework that measures abductive reasoning as a structural inverse problem. Given a reference formal system, such as the lambda-calculus, and a potentially mutated counterpart, agents must determine whether a mutation has occurred and infer the rule modifications responsible for the resulting behavioral differences. Evaluating frontier and mid-tier LLMs reveals a consistent detection-attribution dissociation: models often recognize that a system has been altered but struggle to identify the latent mutations causing the observed discrepancies. Performance degrades substantially under interacting mutations, where models frequently recover only a subset of the underlying mutations. Preliminary evidence also suggests diminishing returns from increased inference-time reasoning, with only modest improvements under larger reasoning budgets, though this finding requires further validation.

Problem

• abductive reasoning
• large language models
• Elenchos
• rule mutations
• behavioral discrepancies

Innovation

• abductive reasoning
• Elenchos
• structural inverse problem
• lambda-calculus
• detection-attribution dissociation