The paper investigates how large language models (LLMs) perform multi‑hop reasoning and challenges the prevailing hop‑aligned circuit hypothesis, which posits that bridge entities are computed sequentially across layers. Through systematic analyses of real‑world multi‑hop queries, the authors discover a phenomenon called layer‑order inversion, where later‑hop answer entities become decodable earlier than bridge entities, and this effect grows with the number of hops. They propose a probabilistic recall‑and‑extract framework that models multi‑hop reasoning as broad probabilistic recall in shallow MLP layers followed by selective extraction in deeper attention layers, and validate this framework with probing analyses that reinterpret prior evidence, explain chain‑of‑thought gains, and diagnose multi‑hop failures.
arXiv:2509. 24653v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) excel at multi-hop reasoning in distribution, yet fail on unseen compositions, a phenomenon known as the curse of two-hop reasoning.
arXiv:2607. 00341v1 Announce Type: cross Abstract: Large language models achieve strong performance on many reasoning tasks when allowed to externalize intermediate steps as Chain-of-Thought (CoT).
Large language models can solve complex multi‑hop tasks but often fail on simple two‑hop queries, even when each hop is individually correct. In a controlled symbolic setting, the authors find that models generalize reliably when the second hop follows the training distribution, but always fail when it deviates. Mechanistic analysis shows that successful generalization relies on consistent intermediate representations across contexts, whereas failures arise from a mismatch between lower‑layer representation construction and upper‑layer mapping to outputs. The study proposes a recurrent‑style training strategy that improves out‑of‑distribution two‑hop generalization.
Large Language Models struggle with implicit multi‑hop reasoning, correctly answering individual facts but failing to combine them in a single pass. In a controlled setting, the authors show that this failure persists even with high 1‑hop accuracy, indicating it is due to pretraining exposure rather than missing knowledge. They test nine data‑centric augmentation formats and find that only individuals seen in compositional contexts during pretraining enable transfer to unseen questions, proving exposure to such contexts is necessary for implicit multi‑hop reasoning.
arXiv:2605.28093v3 Announce Type: replace Abstract: Retrieval-augmented generation (RAG) has emerged as a promising paradigm for enhancing large language models (LLMs) on multi-hop question answering...
Large Language Models fail at implicit multi-hop reasoning: a model answers "When was $X$ born? " and "Who is $Y$'s closest friend?
arXiv:2610.09733v1 Announce Type: new Abstract: Multilingual LLMs answer the same multi-hop reasoning question across languages, but we lack a mechanistic account of whether they share an internal ci...
arXiv:2505.16782v3 Announce Type: replace Abstract: Large Language Models (LLMs) have shown impressive performance on complex tasks through Chain-of-Thought (CoT) reasoning. However, conventional CoT...
arXiv:2607. 10562v1 Announce Type: new Abstract: Evaluating the multi-hop reasoning capabilities of large language models remains a significant challenge.
arXiv:2609.39786v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly combined with knowledge graphs (KGs) to ground reasoning in structured evidence. However, most LLM-based...
The paper introduces MCircKE, a mechanistic circuit-based knowledge editing framework for large language models. MCircKE identifies the causal circuits involved in a specific reasoning task and surgically updates parameters only within those circuits, thereby addressing the reasoning gap where edited facts are not used in multi-step reasoning. Experiments on the MQuAKE-series benchmarks show that this approach improves multi-hop reasoning performance after knowledge editing.
The paper investigates whether different forms of intermediate computation in large language models—such as token-based traces, pause tokens, and latent reasoning—rely on the same underlying mechanism. By training five variants of GPTNeoX on an extended multi-hop reasoning task, the authors find that while vanilla, Chain-of-Thought, and Pause Token models perform well on in-distribution data, they fail to generalize to longer-hop out-of-distribution problems. In contrast, latent-reasoning models exhibit better depth generalization, with causal analysis revealing a sparse recurrent search circuit that implements forward reachability propagation across the graph.