The paper introduces the concept of semantic deception, where unfamiliar symbols are used to mislead large language models (LLMs) during simple arithmetic tasks. By redefining digits and operators with novel symbols, the authors test whether LLMs can abstract and manipulate arbitrary symbol systems or fall back on learned semantic associations. Experiments with four LLMs show that misleading semantic cues significantly degrade performance, revealing a reliance on surface-level semantics that can undermine symbolic reasoning.
Monitoring the chain-of-thought of reasoning artificial intelligence (AI) models remains a key approach to detecting deception and other forms of misbehavior in such models. However, semantic chain-of...
arXiv:2610.10405v1 Announce Type: cross Abstract: Monitoring the chain-of-thought of reasoning artificial intelligence (AI) models remains a key approach to detecting deception and other forms of mis...
The study investigates whether the number of reasoning tokens produced by large language models can signal untruthful responses. Researchers prompted three models to answer 210 multiple‑choice questions under instructions to be truthful, deceptive, or indifferent, and found that truth‑directed answers consistently used fewer reasoning tokens than lie‑directed or indifferent ones. The results suggest that token count could serve as a simple, content‑independent indicator of untruthful behavior when full reasoning traces are unavailable.
The paper introduces OverThink, a slowdown attack that forces reasoning language models (RLMs) to produce many more reasoning tokens while still giving correct answers. By injecting decoy reasoning problems—such as Markov decision processes, language translation, or graphic comprehension—into the model’s context, attackers can dramatically increase token generation (up to 46× on SQuAD and 17× on coding agents). The study evaluates the attack on both proprietary and open-source RLMs across multiple datasets, explores multimodal and coding‑agent variants, and tests several defenses, concluding that defending against OverThink is challenging and that newer RLMs are even more vulnerable due to higher per‑token costs and increased reasoning token usage.
arXiv:2607. 11266v1 Announce Type: new Abstract: Chain-of-Thought (CoT) prompting has significantly advanced the reasoning capabilities of Large Language Models (LLMs), yet it often incurs substantial computational costs due to over-reasoning: the generation of redundant, verbose, or irrelevant steps.
The paper presents a taxonomy of 28 cognitive elements derived from cognitive science to analyze reasoning in large language models (LLMs). Using a fine‑grained evaluation framework, the authors compare 192,000 model traces with 1,346 human think‑aloud traces, finding that while models deploy similar cognitive elements, they differ in usage patterns. The study introduces test‑time reasoning guidance that improves model accuracy on ill‑structured problems by 9.7 percentage points and highlights a research gap in self‑awareness, advocating for a shared vocabulary to diagnose failures and guide robust model development.
arXiv:2509. 14704v3 Announce Type: replace Abstract: Benchmark saturation and training-data contamination increasingly obscure whether reported gains in large language models (LLMs) reflect genuine advances in reasoning or familiarity with recurring patterns in benchmark problems.
The paper investigates whether reasoning representations—explanations for large language model outputs—aid humans in evaluating those outputs. A controlled human study tested six reasoning formats across tasks of varying complexity, measuring structural understanding, error detection, and trust calibration. Results revealed a mismatch: participants favored planning- and decomposition-based representations, yet simpler chain-of-thought traces better supported verification, trust, and interpretability, while preferred formats increased calibration risks.
arXiv:2606. 07410v1 Announce Type: cross Abstract: The emergence of "Aha moments" in large language models, particularly DeepSeek-R1-0120, has raised the question of whether these systems genuinely reason or merely imitate the appearance of reasoning.
arXiv:2607. 08173v1 Announce Type: new Abstract: Black box auditing of language models is an essential pre-deployment tool, but it may miss subtle forms of misalignment and hidden information.
arXiv:2608. 03291v1 Announce Type: cross Abstract: Chain-of-thought (CoT) reasoning improves large language model (LLM) performance while also providing an observable interface to the model's reasoning process.
arXiv:2606. 16934v1 Announce Type: cross Abstract: Reasoning with a Code Interpreter (CI) has emerged as an effective paradigm for enhancing the reasoning capabilities of large language models (LLMs) through executable computation and iterative verification.