Knowledge editing seeks to update language models without full retraining. While most prior work focuses on isolated factual or instance-level edits, we study a more structured setting: mathematical rules as a controlled stress test for abstract symbolic rule editing. We introduce RULE-EDIT, the first benchmark explicitly designed for editing and evaluating rule-level knowledge in LLMs. Our benchmark investigates two key questions: (1) whether edits to abstract rules propagate to derived instances, and (2) whether edits remain consistent across different representations of the same rule. To evaluate these aspects, we propose two new metrics: Instance Portability and Rule Understanding. Experiments on strong open-source LLMs with representative editing methods show that, although models can often overwrite formula-level knowledge, they struggle to transfer edits to rule-derived instances and to maintain consistency across symbolic and natural-language forms. For example, several methods achieve nearly 100% reliability on direct rule queries, while Instance Portability never exceeds 52% and Rule Understanding remains below 26%. Overall, RULE-EDIT provides a controlled benchmark for studying whether current knowledge editing approaches truly modify abstract symbolic rules in LLMs, rather than merely altering surface outputs.

Current Large Language Model (LLM) agents struggle with long-horizon reasoning tasks, such as competition-level mathematics, where sustaining precision over hundreds of steps is critical. A primary bottleneck is the reliance on natural language context or summaries to track progress; these linguistic representations are inherently "lossy," introducing semantic drift and accumulating errors that derail the reasoning chain. To address this, we propose Structured Strategy Folding (SSF), a novel paradigm that shifts the agent’s memory mechanism from linguistic description to computational verification. Instead of compressing history into text, SSF "folds" the reasoning trajectory into a sequence of Executable Symbolic States (ESS). Each ESS serves as a verifiable anchor—comprising either executed code artifacts or formally judged symbolic lemmas—generated through a dual-pathway validation mechanism. By interacting with these rigorous state artifacts rather than raw textual history, the agent effectively eliminates ambiguity and prevents error propagation. Experiments on challenging mathematical benchmarks demonstrate that SSF significantly outperforms standard Chain-of-Thought and existing summarization baselines, achieving substantial accuracy improvements while maintaining robust performance over extended reasoning horizons.

Recent advances in reasoning with large language models (LLMs) have popularized Long Chain-of-Thought (LCoT), a strategy that encourages deliberate and step-by-step reasoning before producing a final answer. While LCoTs have enabled expert-level performance in complex tasks, how the internal structures of their reasoning chains drive, or even predict, the correctness of final answers remains a critical yet underexplored question. In this work, we present LCoT2Tree, an automated framework that converts sequential LCoTs into hierarchical tree structures and thus enables deeper structural analysis of LLM reasoning. Using graph neural networks (GNNs), we reveal that structural patterns extracted by LCoT2Tree, including exploration, backtracking, and verification, serve as stronger predictors of final performance across a wide range of tasks and models. Leveraging an explainability technique, we further identify critical thought patterns such as over-branching that account for failures. Beyond diagnostic insights, the structural patterns by LCoT2Tree support practical applications, including improving Best-of-N decoding effectiveness. Overall, our results underscore the critical role of internal structures of reasoning chains, positioning LCoT2Tree as a powerful tool for diagnosing, interpreting, and improving reasoning in LLMs.

LLMs have marked a revolutonary shift, yet they falter when faced with compositional reasoning tasks. Our research embarks on a quest to uncover the root causes of compositional reasoning failures of LLMs, uncovering that most of them stem from the improperly generated or leveraged implicit reasoning results. Inspired by our empirical findings, we resort to Logit Lens and an intervention experiment to dissect the inner hidden states of LLMs. This deep dive reveals that implicit reasoning results indeed surface within middle layers and play a causative role in shaping the final explicit reasoning results. Our exploration further locates multi-head self-attention (MHSA) modules within these layers, which emerge as the linchpins in accurate generation and leveraing of implicit reasoning results. Grounded on the above findings, we develop CREME, a lightweight method to patch errors in compositional reasoning via editing the located MHSA modules. Our empirical evidence stands testament to CREME’s effectiveness, paving the way for autonomously and continuously enhancing compositional reasoning capabilities in language models.

Large Language Models (LLMs) have recently revolutionized the NLP field, while they still fall short in some specific down-stream tasks. In the work, we focus on utilizing LLMs to perform machine translation, where we observe that two patterns of errors frequently occur and drastically affect the translation quality: language mismatch and repetition. The work sets out to explore the potential for mitigating these two issues by leveraging model editing methods, e.g., by locating Feed-Forward Network (FFN) neurons or something that are responsible for the errors and deactivating them in the inference time.We find that directly applying such methods either limited effect on the targeted errors or has significant negative side-effect on the general translation quality, indicating that the located components may also be crucial for ensuring machine translation with LLMs on the rails.To this end, we propose to refine the located components by fetching the intersection of the locating results under different language settings, filtering out the aforementioned information that is irrelevant to targeted errors. The experiment results empirically demonstrate that our methods can effectively reduce the language mismatch and repetition ratios and meanwhile enhance or keep the general translation quality in most cases.