Emotional intelligence (EI) in large language models (LLMs) is often evaluated through static understanding tasks or single-response dialogue generation. However, emotion management is interactive. A good response should improve the user’s emotional and relational state over several turns. We introduce EIBench, a simulator-based benchmark for interactive emotion management. EIBench contains 2,222 scenarios, with 2,009 for training and 213 for held-out testing. The scenarios are organized by a 2×2 taxonomy covering Support, Defense, Repair, and Charm. We evaluate 15 open- and closed-source LLMs. Current models perform well on support and rapport-building scenes, but struggle with boundary maintenance under user pressure. To improve the EI ability of LLMs, we propose Centered Turn-Credit GRPO (CTC-GRPO), a GRPO extension that reuses the simulator’s per-turn state updates as dense turn-level feedback while preserving the final outcome reward. CTC-GRPO improves Qwen3-8B from -22.4 to +22.4 on EIBench and also improves on out-of-distribution evaluations including SAGE (+12.4) and EQBench3 (+20.9%). Our results show that simulator-tracked user states can support both evaluation and training for multi-turn emotion management.

Large reasoning models (LRMs) have achieved remarkable success on complex tasks, yet their tendency to "overthink” leads to inefficiencies. Although "save-thinking” prompts are intended to mitigate this issue, we find that LRMs still frequently enter the "Still-thinking” mode instead of the expected "No-thinking” mode, especially when facing difficult queries. First, we examine the LRM’s confidence when terminating thinking process, finding that high perplexity signals internal uncertainty and precedes the thinking process. Second, we analyze the divergence in internal attention distributions across different LRMs. Third, we investigate where the LRM allocates its attention within the prompt, identifying a critical "re-reading” pattern. Based on these analyses, we propose an attention intervention method to regulate this behavior. While this intervention successfully suppresses thinking, we find it causes a drop in accuracy, confirming that the thinking process is functional. Our work provides a mechanistic explanation for this behavior, highlighting the trade-off between instruction following and reasoning correctness, and offers insights for developing more controllable LRMs.

Memory management is vital for LLM agents in long-term and personalized interactions.Most previous work studies how to retrieve and use memory, but pays less attention to how memory is extracted.We find two main limitations in existing methods.First, extraction is “ahead-of-time”: the agent saves information before it knows future tasks.A single summary prompt often mixes details, events, and relations, so useful information is lost.Second, extraction is usually one-off. Without verification, errors and hallucinations may stay in memory for a long time. To address these limitations, we propose ProMem, a proactive memory extraction framework.It separates details, events, and relations, and uses different extraction strategies for each type.It also checks completeness to recover missed events and verifies facts at the atomic level to reduce hallucinations.Experiments show that ProMem improves memory completeness and QA accuracy, while keeping a good balance between quality and token cost.

Large language model (LLM) agents increasingly depend on external memory to support long-horizon interaction. Yet many current memory systems still rely on largely fixed default rules once their pipelines are configured. They often apply the same extraction, retrieval, and evidence-organization choices across inputs. This is restrictive. Different memory needs require different choices. We propose AdaMemPIP, an adaptive memory orchestration framework for LLM agents. AdaMemPIP uses an LLM to route both memory storage and memory usage. For storage, it selects extractors for each new batch of dialogue records and writes the extracted items to matched memory libraries. For usage, it selects retrieval methods, target libraries, and organization strategies for each query, then executes the resulting pipeline. The framework is also extensible: new extractors, retrievers, stores, and organizers can be added without changing the overall control logic. This makes AdaMemPIP a flexible framework to long-term agent memory.

Thinking LLMs produce reasoning traces before answering. Prior activation steering work mainly targets on shaping these traces. It remains less understood how answer tokens actually read and integrate the reasoning to produce reliable outcomes. Focusing on quantitative reasoning, we analyze the answer-to-reasoning attention and observe a benign self-reading pattern aligned with correctness, characterized by a forward drift of the reading focus along the reasoning trace and a persistent concentration on key semantic anchors, whereas incorrect solutions exhibit diffuse and irregular attention pattern. We interpret this as internal certainty during answer decoding, where the model commits to a viable solution branch and integrates key evidence. Following this, we propose a training-free steering method driven by Self-Reading Quality (SRQ) scores combining geometric metrics for process control with semantic metrics for content monitoring. SRQ selects data to build steering vectors that guide inference toward benign self-reading and away from uncertain and disorganized reading. Experiments show that our method yields consistent accuracy gains.

Large Language Models (LLMs) are increasingly used for code editing, yet the prevalent full-code generation paradigm suffers from severe efficiency bottlenecks, posing challenges for interactive coding assistants that demand low latency and cost. Despite the predominant focus on scaling model capabilities, the edit format itself has been largely overlooked in model training. In this paper, we begin with a systematic study of conventional diff formats and reveal that fragile offsets and fragmented hunks make generation highly unnatural for LLMs. To address it, we introduce BlockDiff and FuncDiff, two structure-aware diff formats that represent changes as block-level rewrites of syntactically coherent units such as control structures and functions. Furthermore, we propose AdaEdit, a general adaptive edit strategy that trains LLMs to dynamically choose the most token-efficient format between a given diff format and full code. Extensive experiments demonstrate that AdaEdit paired with structure-aware diff formats consistently matches the accuracy of full-code generation, while reducing both latency and cost by over 30% on long-code editing tasks.

Long-horizon tool-use agents need not only to search and plan, but also to decide when to finalize. We study late-stage pressure states, in which an agent is biased toward submitting a final answer that appears complete and polished while key constraints remain unresolved. We first train a linear probe to show that this pressure state is identifiable from the agent’s hidden states. Then, we use activation interventions along this pressure direction and find that shifting the hidden states changes both the pressure score and whether the agent continues tool use or submits early. Through controlled context manipulations, we further see that the pressure is mitigated by constraint clarity and action mapping. Based on these findings, we propose Probe-Sensed Pressure Relief (PSPR), a plugin that applies lightweight pressure relief direction under moderate pressure and moves to structured organization under high pressure risk. Experiments on multiple long-horizon benchmarks show that our method consistently strengthens existing agent methods.

Reinforcement learning (RL) for search agents typically relies on outcome rewards. However, it often fails to achieve effective credit assignment, due to the unclear value of intermediate steps. It is hard to separate their contributions from the final result. In this paper, we propose a dense process supervision method based on fact utility estimation, which models the reasoning process as the accumulation of discrete evidence facts. We first extract structured facts from raw observations and organize them into an explicit fact store. To support credit assignment, we then cluster semantically equivalent facts and infer the posterior utility of each fact cluster using Bayesian estimation over group rollouts. Finally, we convert the estimated fact utilities into dense step-level rewards to guide RL training. Experiments on seven single-hop and multi-hop QA benchmarks show that our method consistently outperforms existing baselines. Ablation studies validate clear relative improvements on multi-hop QA compared to outcome reward-only training.

E-commerce search systems typically separate query-item relevance matching from downstream personalized ranking. While effective for large-scale serving, this design leaves relevance levels dominated by objective query-item matching, forcing follow-up ranking to operate over candidate structures that may not reflect user behavioral preference. We define a new personalized behavioral relevance task, which predicts whether a relevance-stage candidate item is supported by the user’s behavioral evidence under current query intent. We develop a corresponding framework, BRIDGE, for Behavioral Relevance Inference with Dynamic Guidance from Evidence. It organizes multi-timescale behavioral context from online logs, obtains reliable supervision with multi-perspective annotation and adjudication, and trains an LLM to produce structured evidence-grounded decision rationales. It particularly incorporates sample-specific rubric rewards into GRPO to improve generated reasoning traces. Offline experiments validate that BRIDGE improves both prediction accuracy and reasoning quality. Online A/B tests show gains of 0.19% in active user-days and 0.67% in transaction-related user-days, both significant in statistics.

Code translation across multiple programming languages is essential yet challenging due to two vital obstacles: scarcity of parallel data paired with executable test oracles, and optimization imbalance when handling diverse language pairs. We propose BootTrans, a bootstrapping method that resolves both obstacles. Its key idea is to leverage the functional invariance and cross-lingual portability of test suites, adapting abundant pivot-language unit tests to serve as universal verification oracles for multilingual reinforcement learning (RL) training. Our method introduces a dual-pool architecture with seed and exploration pools to progressively expand training data via execution-guided experience collection.Furthermore, we design a language-aware weighting mechanism that dynamically prioritizes harder translation directions based on relative performance across sibling languages, mitigating optimization imbalance. Extensive experiments on the HumanEval-X and TransCoder-Test benchmarks demonstrate substantial improvements over baseline LLMs across all translation directions, with ablation studies validating the effectiveness of both bootstrapping and weighting components.

Retrieval-augmented generation (RAG) has emerged as a promising technology for addressing hallucination issues in the responses generated by large language models (LLMs). Existing studies on RAG primarily focus on applying semantic-based approaches to retrieve isolated relevant chunks, which ignore their intrinsic relationships. In this paper, we propose a novel Knowledge Graph-Guided Retrieval Augmented Generation (KG²RAG) framework that utilizes knowledge graphs (KGs) to provide fact-level relationships between chunks, improving the diversity and coherence of the retrieved results. Specifically, after performing a semantic-based retrieval to provide seed chunks, KG²RAG employs a KG-guided chunk expansion process and a KG-based chunk organization process to deliver relevant and important knowledge in well-organized paragraphs. Extensive experiments conducted on the HotpotQA dataset and its variants demonstrate the advantages of KG²RAG compared to existing RAG-based approaches, in terms of both response quality and retrieval quality.

In a rapidly evolving world where information updates swiftly, knowledge in large language models (LLMs) becomes outdated quickly. Retraining LLMs is not a cost-effective option, making knowledge editing (KE) without modifying parameters particularly necessary. We find that although existing retrieval-augmented generation (RAG)-based KE methods excel at editing simple knowledge, they struggle with KE in multi-hop question answering due to the issue of ”edit skipping”, which refers to skipping the relevant edited fact in inference. In addition to the diversity of natural language expressions of knowledge, edit skipping also arises from the mismatch between the granularity of LLMs in problem-solving and the facts in the edited memory. To address this issue, we propose a novel Iterative Retrieval-Augmented Knowledge Editing method with guided decomposition (IRAKE) through the guidance from single edited facts and entire edited cases. Experimental results demonstrate that IRAKE mitigates the failure of editing caused by edit skipping and outperforms state-of-the-art methods for KE in multi-hop question answering.

We study continual information extraction (IE), which aims to extract emerging information across diverse IE tasks incessantly while avoiding forgetting. Existing approaches are either task-specialized for a single IE task or suffer from catastrophic forgetting and insufficient knowledge transfer in continual IE. This paper proposes a new continual IE model using token-level mixture of LoRA experts with LLMs. We leverage a LoRA router to route each token to the most relevant LoRA experts, facilitating effective knowledge transfer among IE tasks. We guide task experts’ selection by task keys to retain the IE task-specific knowledge and mitigate catastrophic forgetting. We design a gate reflection method based on knowledge distillation to address forgetting in the LoRA router and task keys. The experimental results show that our model achieves state-of-the-art performance, effectively mitigating catastrophic forgetting and enhancing knowledge transfer in continual IE.

In this paper, we identify a critical problem, “lost-in-retrieval”, in retrieval-augmented multi-hop question answering (QA): the key entities are missed in LLMs’ sub-question decomposition. “Lost-in-retrieval” significantly degrades the retrieval performance, which disrupts the reasoning chain and leads to the incorrect answers. To resolve this problem, we propose a progressive retrieval and rewriting method, namely ChainRAG, which sequentially handles each sub-question by completing missing key entities and retrieving relevant sentences from a sentence graph for answer generation. Each step in our retrieval and rewriting process builds upon the previous one, creating a seamless chain that leads to accurate retrieval and answers. Finally, all retrieved sentences and sub-question answers are integrated to generate a comprehensive answer to the original question. We evaluate ChainRAG on three multi-hop QA datasets—MuSiQue, 2Wiki, and HotpotQA—using three large language models: GPT4o-mini, Qwen2.5-72B, and GLM-4-Plus. Empirical results demonstrate that ChainRAG consistently outperforms baselines in both effectiveness and efficiency.

Continual few-shot relation extraction (RE) aims to continuously train a model for new relations with few labeled training data, of which the major challenges are the catastrophic forgetting of old relations and the overfitting caused by data sparsity. In this paper, we propose a new model, namely SCKD, to accomplish the continual few-shot RE task. Specifically, we design serial knowledge distillation to preserve the prior knowledge from previous models and conduct contrastive learning with pseudo samples to keep the representations of samples in different relations sufficiently distinguishable. Our experiments on two benchmark datasets validate the effectiveness of SCKD for continual few-shot RE and its superiority in knowledge transfer and memory utilization over state-of-the-art models.

Continual relation extraction (RE) aims to learn constantly emerging relations while avoiding forgetting the learned relations. Existing works store a small number of typical samples to re-train the model for alleviating forgetting. However, repeatedly replaying these samples may cause the overfitting problem. We conduct an empirical study on existing works and observe that their performance is severely affected by analogous relations. To address this issue, we propose a novel continual extraction model for analogous relations. Specifically, we design memory-insensitive relation prototypes and memory augmentation to overcome the overfitting problem. We also introduce integrated training and focal knowledge distillation to enhance the performance on analogous relations. Experimental results show the superiority of our model and demonstrate its effectiveness in distinguishing analogous relations and overcoming overfitting.

Relation linking (RL) is a vital module in knowledge-based question answering (KBQA) systems. It aims to link the relations expressed in natural language (NL) to the corresponding ones in knowledge graph (KG). Existing methods mainly rely on the textual similarities between NL and KG to build relation links. Due to the ambiguity of NL and the incompleteness of KG, many relations in NL are implicitly expressed, and may not link to a single relation in KG, which challenges the current methods. In this paper, we propose an implicit RL method called ImRL, which links relation phrases in NL to relation paths in KG. To find proper relation paths, we propose a novel path ranking model that aligns not only textual information in the word embedding space but also structural information in the KG embedding space between relation phrases in NL and relation paths in KG. Besides, we leverage a gated mechanism with attention to inject prior knowledge from external paraphrase dictionaries to address the relation phrases with vague meaning. Our experiments on two benchmark and a newly-created datasets show that ImRL significantly outperforms several state-of-the-art methods, especially for implicit RL.