Multi-hop QA benchmarks often reward Large Language Models (LLMs) for spurious correctness, where models reach correct answers through invalid intermediate reasoning. We propose SAFE, an LLM-as-verifier framework for evidence-grounded multi-hop QA. Rather than judging only the final answer after generation, SAFE verifies reasoning during generation by checking intermediate steps against the provided passages and previous reasoning trajectory. To make this process checkable, SAFE decomposes reasoning into atomic, evidence-grounded units represented with Knowledge Graph (KG) triples. At train-time, SAFE verifies benchmark supervision under KG-grounded constraints and constructs reliable verifier training data. At inference-time, an external verifier checks each generated step, identifies invalid reasoning, and provides correction feedback before errors propagate. Across three multi-hop QA benchmarks, SAFE improves accuracy by 8.8 pp on average. These results show that evidence-grounded multi-hop QA benefits from shifting LLM-based evaluation from post-hoc answer judgment to stepwise reasoning verification.
Recent advances in large language models (LLMs) have shown promise in formal theorem proving, yet evaluating semantic correctness remains challenging. Existing evaluations rely on indirect proxies such as lexical overlap with human-annotated proof,or expensive manual inspection.Inspired by the shift from lexical comparison to test-based evaluation in code generation, we propose T^2, a framework that evaluates the semantic correctness of formal theorems: a generated theorem is considered correct only if all dependent successor theorems compile successfully, analogous to integration testing.We construct a benchmark from 5 real-world Lean 4 repositories, comprising 2,206 problems paired with 41 successor theorems on average, automatically extracted without human effort.Experiments demonstrate that while state-of-the-art models achieve high compilation success, they perform significantly worse under our semantic metric.The best model, Claude-Sonnet-4.5, achieves only 38.9% Testing Accuracy on the full set, given both natural language proof and successor theorems as context, revealing a critical gap in current theorem generation capabilities.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track) · 2026 · 正式发表
Dohyeon Lee Korea Advanced Institute of Science & Technology, Yeonseok Jeong Seoul National University, Seung-won Hwang Seoul National University
Reasoning-intensive information retrieval uses large language models to solve complex queries via multi-step reasoning. However, existing methods have critical limitations. Chain-of-Thought (CoT) approaches suffer from inefficiency, while state-based methods, despite better token efficiency, often fall into reasoning cycles that trap the query refinement process. To address these issues, we propose Episodic Memory for Retrieval (EMR), which enhances the state-based framework with an episodic memory. This module stores the full history of prior states for a query, allowing the model to avoid repetition of such cycles. Experiments on the BRIGHT benchmark show that EMR consistently outperforms both CoT and state-based baselines. Moreover, it is highly token-efficient, reducing token usage by 72% on average. Our results show that episodic memory is an effective and token-efficient mechanism for reasoning-intensive retrieval. The gains also generalize across different base models and stay efficient in terms of end-to-end latency. The code is available in https://github.com/ldilab/EMR.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Jongho Kim Seoul National University, Jaeyoung Kim, Jihyuk Kim LG Corporation, Yu Jin Kim LG AI Research, Seung-won Hwang Seoul National University, Moontae Lee LG Corporation and University of Illinois, Chicago
We study leveraging adaptive retrieval to ensure sufficient “bridge” documents are retrieved for reasoning-intensive retrieval. Bridge documents are those that contribute to the reasoning process yet are not directly relevant to the initial query. While existing reasoning-based reranker pipelines attempt to surface these documents in ranking, they suffer from bounded recall. Naive solution with adaptive retrieval into these pipelines often leads to planning error propagation. To address this, we propose REPAIR, a framework that bridges this gap by repurposing reasoning plans as dense feedback signals for adaptive retrieval. Our key distinction is enabling mid-course correction during reranking through selective adaptive retrieval, retrieving documents that support the pivotal plan. Experimental results on reasoning-intensive retrieval and complex QA tasks demonstrate that our method outperforms existing baselines by 5.6%pt.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Youngwon Lee Seoul National University, Seung-won Hwang Seoul National University, Zhewei Yao Snowflake, Yuxiong He Microsoft
Retrieval-augmented generation needs generation to follow retrieved evidence across shifting domains and prompt layouts, but training a new stronger model per task is costly. To this end, we propose GRAD, an adaptive decoding-time framework that keeps the base generator fixed and composes small, objective-specific guidance at inference. A key advantage of this design is enabling mix and match diverse RAG objectives: model scaling (MS), domain adaptation (DA) and positional debiasing (DB) can be integrated as token-level guidance terms, and new objectives can be easily plugged in. Across public benchmarks and private settings with no in-domain labels, GRAD improves accuracy with favorable latency, offering strong trade-offs versus scaling while reliably activating helpful objectives and suppressing harmful ones, adaptively to tasks.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表