Deep Research Agents (DR-Agents) extend LLMs with multi-step retrieval, tool use, and intermediate computation to solve complex problems. However, their execution logs often contain redundant context, noisy evidence, and brittle dependencies among intermediate decisions. As a result, final answers can be sensitive to irrelevant or weakly verified information. We propose DualResearch, a log-based post-processing framework for answer reconstruction. Given a query and a completed DR-Agent log, DualResearch normalizes the log into event records and constructs two evidence graphs: a breadth semantic graph for declarative evidence and a depth causal graph for procedural provenance. It then retrieves breadth nodes and depth paths, assigns higher weights to retrieval channels with more concentrated retained candidate distributions, and jointly considers relevance and cross-view agreement to select evidence for reconstruction; the concentration measure reflects retrieval confidence rather than factual correctness. Experiments on HLE, GAIA, and GPQA show consistent accuracy improvements over underlying DR-Agents and controlled log-reuse baselines.

Online lifelong learning agents must decide not only how to act but also when to consult prior experience to continually improve on long-horizon tasks. Existing methods typically retrieve memories passively, such as at task initialization or after each step, and therefore miss knowledge gaps that arise during interaction. We propose ProactAgent, an experience-driven lifelong learning framework for proactive retrieval over a structured Experience Base. ProactAgent continually improves through ExpOnEvo, which jointly updates policies and refines memory, organizing past interactions into factual, episodic, and skill repositories. It further introduces ProactRL, which treats retrieval as an explicit policy action and learns when and what to retrieve. By comparing paired continuations from identical interaction prefixes with and without retrieval, ProactRL provides step-level process rewards that encourage retrieval only when it improves task outcomes or efficiency. Experiments on SciWorld, AlfWorld, and StuLife show that ProactAgent consistently outperforms all baselines, achieving up to 32% relative improvement in success rate and over 33% reduction in interaction rounds.

As AI research increasingly moves beyond systems optimized for static tasks, a central challenge is building agents that learn continuously through experience-driven lifelong learning (ELL). In this paper, we introduce StuLife, a synthetic benchmark designed to evaluate selected operational capabilities relevant to ELL. Particularly, StuLife simulates a student’s college journey, from enrollment to academic and personal development, across three core phases and ten detailed sub-scenarios. StuLife is designed around four key design principles: From Isolated Tasks to Stateful Interaction, From Passive to Proactive, From Context to Memory, and From Imitation to Learning. In this dynamic environment, agents must acquire and distill practical skills and maintain persistent memory to make decisions based on evolving state variables (e.g., resource availability and time), and are also expected to initiate actions without external prompting. To this end, StuLife provides an evaluation protocol with dedicated metrics (e.g., StuGPA) for these capabilities. Our evaluation reveals that even the highest-StuGPA model, GPT-5, scores only 17.9/100, highlighting large gaps in long-term retention and proactive task initiation. Beyond evaluating LLMs on StuLife, we also explore the role of context engineering in improving agent performance on lifelong learning tasks.

Existing continual knowledge editing methods suffer from cognitive dissonance: after an edit, the model answers the edit prompt correctly but reverts to the old fact under complex equivalent queries. This means old and new facts coexist rather than the old being replaced. The root cause of cognitive dissonance is that existing methods fit the new fact on the target prompt without overwriting the old one. In constrained methods, static null-space projectors lock outdated representations into the protected subspace, forcing updates into orthogonal directions. To resolve cognitive dissonance, we propose P-Space (Puncturing the Null-Space to Resolve Cognitive Dissonance in Continual Knowledge Editing), which at each edit step (1) punctures the protected covariance to unprotect the outdated representation, (2) overwrites the old mapping in closed form along its original direction under the punctured projector, and (3) anchors the new fact into the covariance with adaptive scaling so that future edits preserve it. Experiments on two datasets show that P-Space outperforms existing methods while preserving general capabilities.

Large language models (LLMs) are increasingly used in role-playing applications, yet evaluating whether they maintain character-consistent behavior over extended interactions remains challenging. Existing benchmarks mostly focus on short exchanges, static long-context inputs, or isolated aspects of role-playing, making long-horizon failures difficult to diagnose. We present REMP (Relationship-Emotion-Memory-Persona), a structured event-probing benchmark for long-horizon role-playing dialogue. REMP constructs multi-chapter dialogue trajectories from character profiles and embeds diagnostic probe events targeting persona consistency, emotional consistency, memory integrity, and relational continuity. We further introduce REMP-EVAL, an anchor-based multi-turn evaluation protocol that assesses model behavior after event triggers using dimension-specific references. The current release contains 160 characters organized into 80 dialogue trajectories, each averaging about 60 sessions and 800 turns. Experiments on 11 LLMs show that memory integrity and emotional consistency are the main bottlenecks, indicating that long-horizon role-playing requires not only long-context retrieval but also sustained affective and relational consistency.

Large Vision-Language Models (LVLMs) commonly suffer from hallucinations, which appear as fabricating non-existent objects or outputting incorrect attributes and spatial relationships in images. Previous works empirically indicate that CLIP-style vision encoders tend to focus on primary objects while ignoring image details, which is one of the causes of hallucinations. To address this issue, we propose plug-and-play fine-grained visual experts (PLANE) comprising two components: LoRA visual experts operating on CLIP for multi-aspect image encoding, and an instruction-aware gate network for adaptive image feature fusion. During training, we construct a fine-grained image-caption pairs dataset, where images are described from three aspects: objects, attributes, and relations between objects. Image-caption pairs from different aspects are used to train the three fine-grained LoRA experts respectively. Aiming at comprehensively evaluate our method’s capabilities, we manually annotate a fine-grained hallucination evaluation benchmark for LVLMs which is GPT4-independent during evaluation phase (GIFT). We conduct experiments both on popular hallucination evaluation benchmarks and GIFT. The results show that our method effectively reduces hallucinations across models of different types and sizes while introducing minimal inference latency. Compared to advanced baselines, our model also demonstrates superior performance.

Despite the pivotal role of numerical reasoning as the cornerstone of mathematical capabilities in large language models (LLMs) across applications, few benchmarks evaluate LLMs by integrating numerical processing and mathematical reasoning, hindering the interpretability of failures in math tasks. We introduce PyraMathBench, a comprehensive hierarchical benchmark with 27,215 questions derived from 7,404 math word problems, spanning 4 key cognitive aspects, 14 subcategories, and 2 modalities. Experiments reveal that LLMs’ performance is severely compromised by inadequate numerical computation and weak handling of abstract numerical questions. To address this, we propose the Smart Optimization Learning-based VErsatile module (SOLVE) and Interactive Relative Policy Optimization (IRPO), which enhance LLMs’ numerical-mathematical synergy via efficient tool calls (fuzzy matching and low-quality call rejection). Comparative experiments show Qwen-2.5 achieves a 5.0 score improvement with SOLVE and IRPO training.

Knowledge Tracing (KT) is a pivotal task in personalized education, aiming to predict students’ future performance based on their historical interactions. While prior work has focused on learning behavioral sequences using question IDs or surface-level textual features, these methods often fail to capture complex behavioral patterns due to a lack of deep reasoning capabilities and world knowledge. To address this, we propose LLM-KT, a novel framework that integrates the reasoning power of Large Language Models (LLMs) with the sequential modeling strengths of traditional KT methods via multi-level plug-and-play alignment. Specifically, for task-level alignment, we design a plug-and-play instruction to leverage the rich knowledge and reasoning capacity of LLMs for the KT objective. For modality-level alignment, we introduce two mechanisms to integrate representations learned by traditional methods: (1) a Semantic History Projector that flexibly inserts compressed context embeddings into LLMs using question- and concept-specific tokens to capture long-term history; and (2) a Behavioral Dynamics Projector that enhances LLMs with sequential interaction patterns via a sequence adapter. Extensive experiments on four standard datasets demonstrate that LLM-KT achieves state-of-the-art performance, significantly outperforming over 20 competitive baselines.

To develop a reliable AI for psychological assessment, we introduce PsychEval, a multi-session, multi-therapy, and highly realistic benchmark designed to address three key challenges:1) Can we train a highly realistic AI counselor? Realistic counseling is a longitudinal task requiring sustained memory and dynamic goal tracking. We propose a multi-session benchmark (spanning 6-10 sessions across three distinct stages) that demands critical capabilities such as memory continuity, adaptive reasoning, and longitudinal planning. The dataset is annotated with extensive professional skills, comprising over 677 meta-skills and 4577 atomic skills. 2) How to train a multi-therapy AI counselor? While existing models often focus on a single therapy, complex cases frequently require flexible strategies among various therapies. We construct a diverse dataset covering five therapeutic modalities alongside an integrative therapy with a unified three-stage clinical framework across six core psychological topics. 3) How to systematically evaluate an AI counselor? We establish a holistic evaluation framework with 18 therapy-specific and therapy-shared metrics across Client-Level and Counselor-Level dimensions. To We also construct over 2,000 diverse client profiles. Extensive experimental analysis fully validates the superior quality and clinical fidelity of our dataset.Our datasets and evaluation framework are anonymously available at this repository.

Graph-based retrieval-augmented generation (GraphRAG) is effective for knowledge-intensive and multi-hop query tasks; however, many existing methods primarily seed entity-based graphs and rely on implicit semantic relevance propagation. This often (i) under-retrieves when user queries are abstract and semantically sparse at the entity level, and (ii) suffers from brittle multi-hop reasoning, where noisy activations can derail entity-to-entity transitions and corrupt the inferred relation chain, yielding unreliable conclusions. To this end, we propose FlowRAG, a semantic-aware retrieval framework that improves both semantic recall and explicit reasoning. Specifically, FlowRAG constructs a quad-level heterogeneous graph over passages, summaries, sentences, and entities, where summary nodes serve as a coarse semantic hub. At retrieval time, a dual-granularity activation module combines summary–query alignment with sentence-level matching to activate relevant entities under paraphrase and abstraction robustly. We then introduce a frequency-aware weighted flow module that routes relevance through entity–passage links weighted by within-passage term frequency, pruning noisy connections and extracting high-confidence reasoning paths as an explicit logic skeleton for generation. Extensive experiments show that FlowRAG obtains state-of-the-art performance on complex reasoning benchmarks.

Model merging aims to integrate multiple expert models into a single model that inherits their complementary strengths without incurring the inference-time cost of ensembling. Recent progress has shown that merging can be highly effective when all source models are homogeneous, i.e., derived from the same pretrained backbone and therefore share aligned parameter coordinates or compatible task vectors. Yet this assumption is increasingly unrealistic in open model ecosystems, where useful experts are often built on different families such as Llama, Qwen, and Mistral. In such heterogeneous settings, direct weight-space fusion becomes ill-posed due to architectural mismatch, latent basis misalignment, and amplified cross-source conflict. We address this problem with HeteroFusion for heterogeneous language model fusion, which consists of two key components: topology-based alignment that transfers knowledge across heterogeneous backbones by matching functional module structures instead of raw tensor coordinates, and conflict-aware denoising that suppresses incompatible or noisy transfer signals during fusion, with a small replay set used to assist transfer-operator optimization. We further provide analytical and empirical evidence for preserving the target adapter basis while predicting structured updates. Across heterogeneous transfer, multi-source fusion, noisy-source robustness, and cross-family generalization settings, HeteroFusion consistently outperforms strong merging, fusion, and ensemble baselines.

Continual Learning (CL) models, while adept at sequential knowledge acquisition, face significant and often overlooked privacy challenges due to accumulating diverse information. Traditional privacy methods, like a uniform Differential Privacy (DP) budget, indiscriminately protect all data, leading to substantial model utility degradation and hindering CL deployment in privacy-sensitive areas. To overcome this, we propose a privacy-enhanced continual learning (PeCL) framework that forgets what’s sensitive and remembers what matters. Our approach first introduces a token-level dynamic Differential Privacy strategy that adaptively allocates privacy budgets based on the semantic sensitivity of individual tokens. This ensures robust protection for private entities while minimizing noise injection for non-sensitive, general knowledge. Second, we integrate a privacy-guided memory sculpting module. This module leverages the sensitivity analysis from our dynamic DP mechanism to intelligently forget sensitive information from the model’s memory and parameters, while explicitly preserving the task-invariant historical knowledge crucial for mitigating catastrophic forgetting. Extensive experiments show that PeCL achieves a superior balance between privacy preserving and model utility, outperforming baseline models by maintaining high accuracy on previous tasks while ensuring robust privacy.

Mixture of Experts (MoE) has great potential in scaling up the capacity of large models while maintaining low computational costs. However, existing methods mainly reduce redundancy at the expert level while overlooking fine-grained neuron-level redundancy and expert collaboration. In this paper, we propose an Information Bottleneck based MoE (IBMoE) for parameter-efficient fine-tuning, which reduces neuron-level redundancy within each expert and fosters internal collaboration among all experts. Specifically, IBMoE equips each expert with learnable queries to capture task-specific characteristics and employs sparse neuronal activation to dynamically select the most relevant neurons, reducing redundant activations across tasks. In addition, a diversity constraint is imposed among experts to maximize their knowledge diversity, enabling them to cooperate more effectively. Extensive experiments across multiple benchmarks show that IBMoE consistently outperforms strong baselines while activating significantly fewer parameters. In addition, IBMoE reduces inference latency by 61.7% and the memory footprint of adapters by 48.2% compared with previous MoE-based adaptation methods, while achieving comparable training costs.

Fine-tuning large language models (LLMs) on domain-specific datasets has become a standard paradigm for adapting LLMs to specialized applications. However, recent work has shown that even fine-tuning on benign task-specific data can substantially weaken the safety capabilities of LLMs. While existing efforts have made progress in identifying data responsible for safety degradation, they usually rely on a single mean vector computed over a specific model with its tokenizer to represent the safety direction, which limits both the effectiveness and transferability of their risk assessment measures. To address these limitations, we propose DataShield, a data assessment framework that identifies risky fine-tuning samples and response segments through consensus subspace alignment over joint safety-critical semantic spaces derived from multiple safety-aligned LLMs.Within these spaces, DataShield extracts consensus safe and unsafe subspaces using semantic spectral decomposition over safe and unsafe data representations. The risk of a data sample or segment is then estimated by measuring its relative alignment with the unsafe and safe subspaces, enabling both sample-level filtering and fine-grained segment-level masking. Compared with state-of-the-art filtering and masking baselines, DataShield reduces ASR by 14.6% with sample filtering and 32.3% with segment masking, while preserving downstream utility and avoiding target-model-specific risk computation.

Long-horizon autonomous agents require memory systems to retain historical information, track evolving states, and reuse relevant knowledge beyond finite context windows. Existing agentic memory systems typically follow a memory construction–retrieval (MCR) pipeline, but often adapt mainly the memory bank while keeping the surrounding pipeline fixed after deployment. This fixed-pipeline design struggles to handle heterogeneous task-specific failure modes and can become misaligned with memory banks that evolve in scale and structure over time. To address these limitations, we propose MemPro, a system-level evolution framework that treats the entire MCR pipeline as an evolvable program rather than adapting only the memory bank or prompt text. MemPro maintains a version tree of runnable memory-system implementations, where an Evolving Agent iteratively selects promising versions, diagnoses recurring failures, and creates improved child versions through failure-mode-guided edit–debug refinement. Experiments on LongMemEval, LoCoMo, HotpotQA, and NarrativeQA show that MemPro consistently outperforms strong static and prompt-level evolving baselines within a few iterations, continues to improve with evolution, and achieves a favorable performance–cost trade-off. Code is available at https://github.com/wanghai673/MemPro.

Large reasoning models achieve strong performance on complex tasks by generating extended chain-of-thought (CoT) traces via reinforcement learning with verifiable rewards (RLVR). While current RLVR methods have achieved strong results with correctness-based reward signals, they provide limited guidance on the quality of the reasoning process itself, leaving the internal reasoning structure largely unoptimized. Through empirical analysis across multiple model families, we identify a consistent pattern: correct reasoning traces exhibit more frequent and larger token-level entropy drops within the thinking phase than incorrect ones. We propose ERR+, a two-phase RLVR framework grounded in this observation. The first phase trains with the Entropy Relief Reward (ERR), a bonus proportional to cumulative token-level entropy drops in the thinking phase, log-normalized by response length. Unlike prior methods that suppress entropy, ERR rewards the resolution of uncertainty while leaving exploratory high-entropy states unconstrained. The second phase introduces the Robust Relative Efficiency Reward, which scores each response’s length against co-generated peers via a tanh-transformed within-group z-score. We provide a formal analysis showing that joint optimization of the two objectives induces gradient conflict in early training, motivating the sequential design (Theorem 1). Experiments on five datasets demonstrate consistent improvements in both accuracy and response conciseness across model backbones.Our code is available at https://github.com/XrkArul/err_response.

Knowledge editing is a promising method for updating Large Language Models efficiently. However, previous studies often suffer from poor specificity in continual editing, as they typically focus on single edits or preventing knowledge forgetting. To address this, we propose TamEdit, a trajectory-aware meta-learning method that preserves specificity for continual knowledge editing. TamEdit unifies three levels: Inner Optimization performs multi-step fast fine-tuning on the single edit; Trajectory-based Editing unifies continual edits with a growing memory; and Outer Optimization leverages meta-learning to distill cross-task strategies for preserving specificity. By capturing the relationships between different single edits within the trajectory, our method learns how to effectively avoid specificity drift. Experiments across multiple LLMs show TamEdit significantly outperforms baselines in continual editing, improving specificity by 14.81% with fast speed (requiring only 8.84% of the time cost of most baselines), while preserving general capabilities.

Reasoning is an important task for large language models (LLMs). Among all the reasoning paradigms, inductive reasoning is one of the basic types, which is characterized by its particular-to-general thinking process and the non-uniqueness of its answers. The inductive mode is crucial for knowledge generalization and aligns better with human cognition, so it is a fundamental mode of learning, hence attracting increasing interest. Despite the importance of inductive reasoning, there is no systematic summary of it. Therefore, this paper presents the first comprehensive survey of inductive reasoning for LLMs. First, methods for improving inductive reasoning are categorized into three main areas: post-training enhancement, test-time exploration, and data augmentation. Then, current benchmarks of inductive reasoning are summarized, and a unified sandbox-based evaluation approach with the observation coverage metric is derived. Finally, we offer some analyses regarding the source of inductive ability and how simple model architectures and data help with inductive tasks, providing a solid foundation for future research.

Personalized large language models (LLMs) have attracted great attention in many applications, such as emotional support and role-playing. However, existing works primarily focus on modeling explicit character profiles, while ignoring the underlying personality traits that truly shape behaviors and decision-making, hampering the development of more anthropomorphic and psychologically-grounded AI systems. In this paper, we explore the modeling of Big Five personality traits, which is the most widely used trait theory in psychology, and propose P-React, a mixture of experts (MoE)-based personalized LLM. Particularly, we integrate a Personality Specialization Loss (PSL) to better capture individual trait expressions, providing a more nuanced and psychologically grounded personality simulacrum. To facilitate research in this field, we curate OCEAN-Chat, a high-quality, human-verified dataset designed to train LLMs in expressing personality traits across diverse topics. Extensive experiments demonstrate the effectiveness of P-React in maintaining consistent and real personality.

Retrieval-augmented generation (RAG) is usually integrated into large language models (LLMs) to mitigate hallucinations and knowledge obsolescence. Whereas, conventional one-step retrieve-and-read methods are insufficient for multi-hop question answering, facing challenges of retrieval semantic mismatching and the high cost in handling interdependent subquestions. In this paper, we propose Optimizing Question Semantic Space for Dynamic Retrieval-Augmented Multi-hop Question Answering (Q-DREAM). Q-DREAM consists of three key modules: (1) the Question Decomposition Module (QDM), which decomposes multi-hop questions into fine-grained subquestions; (2) the Subquestion Dependency Optimizer Module (SDOM), which models the interdependent relations of subquestions for better understanding; and (3) the Dynamic Passage Retrieval Module (DPRM), which aligns subquestions with relevant passages by optimizing the semantic embeddings.Experimental results across various benchmarks demonstrate that Q-DREAM significantly outperforms existing RAG methods, achieving state-of-the-art performance in both in-domain and out-of-domain settings. Notably, Q-DREAM also improves retrieval efficiency while maintaining high accuracy compared with recent baselines.

With the proliferation of large language models (LLMs) in the medical domain, there is increasing demand for improved evaluation techniques to assess their capabilities. However, traditional metrics like F1 and ROUGE, which rely on token overlaps to measure quality, significantly overlook the importance of medical terminology. While human evaluation tends to be more reliable, it can be very costly and may as well suffer from inaccuracies due to limits in human expertise and motivation. Although there are some evaluation methods based on LLMs, their usability in the medical field is limited due to their proprietary nature or lack of expertise. To tackle these challenges, we present AutoMedEval, an open-sourced automatic evaluation model with 13B parameters specifically engineered to measure the question-answering proficiency of medical LLMs. The overarching objective of AutoMedEval is to assess the quality of responses produced by diverse models, aspiring to significantly reduce the dependence on human evaluation. Specifically, we propose a hierarchical training method involving curriculum instruction tuning and an iterative knowledge introspection mechanism, enabling AutoMedEval to acquire professional medical assessment capabilities with limited instructional data. Human evaluations indicate that AutoMedEval surpasses other baselines in terms of correlation with human judgments.

Prevailing research concentrates on superficial features or descriptions of images, revealing a significant gap in the systematic exploration of their connotative and aesthetic attributes. Furthermore, the use of cross-modal relation detection modules to eliminate noise from comprehensive image representations leads to the omission of subtle contextual information. In this paper, we present a Visual Connotation and Aesthetic Attributes Understanding Network (Vanessa) for Multimodal Aspect-based Sentiment Analysis. Concretely, Vanessa incorporates a Multi-Aesthetic Attributes Aggregation (MA3) module that models intra- and inter-dependencies among bi-modal representations as well as emotion-laden aesthetic attributes. Moreover, we devise a self-supervised contrastive learning framework to explore the pairwise relevance between images and text via the Gaussian distribution of their CLIP scores. By dynamically clustering and merging multi-modal tokens, Vanessa effectively captures both implicit and explicit sentimental cues. Extensive experiments on widely adopted two benchmarks verify Vanessa’s effectiveness.

Though large language models (LLMs) achieve significant success in recent years, the hallucination issue remains a challenge, and numerous benchmarks are proposed for hallucination detection. Nevertheless, some of these benchmarks are not naturally generated by LLMs but are intentionally induced. Also, many merely focus on the factuality hallucination while ignoring the faithfulness hallucination. Additionally, although dialogue pattern is more widely utilized in the era of LLMs, current benchmarks only concentrate on sentence-level and passage-level hallucination. In this study, we propose DiaHalu, the first dedicated dialogue-level hallucination evaluation benchmark for LLMs to our knowledge. Initially, we integrate the collected topics into system prompts and facilitate a dialogue between two LLMs. Subsequently, we manually modify the contents that do not adhere to human language conventions and then have LLMs re-generate, simulating authentic human-machine interaction scenarios. Finally, professional scholars annotate all the samples in the dataset. DiaHalu covers four common multi-turn dialogue domains and five hallucination subtypes, extended from factuality and faithfulness hallucination. Experiments through some well-known LLMs and detection methods on the dataset show that DiaHalu is a challenging benchmark, holding significant value for further research.

Aspect-based sentiment analysis (ABSA) is an important subtask of sentiment analysis, which aims to extract the aspects and predict their sentiments. Most existing studies focus on improving the performance of the target domain by fine-tuning domain-specific models (trained on source domains) based on the target domain dataset. Few works propose continual learning tasks for ABSA, which aim to learn the target domain’s ability while maintaining the history domains’ abilities. In this paper, we propose a Large Language Model-based Continual Learning (LLM-CL) model for ABSA. First, we design a domain knowledge decoupling module to learn a domain-invariant adapter and separate domain-variant adapters dependently with an orthogonal constraint. Then, we introduce a domain knowledge warmup strategy to align the representation between domain-invariant and domain-variant knowledge. In the test phase, we index the corresponding domain-variant knowledge via domain positioning to not require each sample’s domain ID. Extensive experiments over 19 datasets indicate that our LLM-CL model obtains new state-of-the-art performance.

Emotion-cause pair extraction (ECPE) aims to identify emotion clauses and their corresponding cause clauses within a document. Traditional methods often rely on coarse-grained clause-level annotations, which can overlook valuable fine-grained clues. To address this issue, we propose Multi-Granularity Clue Learning (MGCL), a novel approach designed to capture fine-grained emotion-cause clues from a weakly-supervised perspective efficiently. In MGCL, a teacher model is leveraged to give sub-clause clues without needing fine-grained annotated labels and guides a student model to identify clause-level emotion-cause pairs. Furthermore, we explore domain-invariant extra-clause clues under the teacher model’s advice to enhance the learning process. Experimental results on the benchmark dataset demonstrate that our method achieves state-of-the-art performance while offering improved interpretability.

With the proliferation of Large Language Models (LLMs) in diverse domains, there is a particular need for unified evaluation standards in clinical medical scenarios, where models need to be examined very thoroughly. We present CliMedBench, a comprehensive benchmark with 14 expert-guided core clinical scenarios specifically designed to assess the medical ability of LLMs across 7 pivot dimensions. It comprises 33,735 questions derived from real-world medical reports of top-tier tertiary hospitals and authentic examination exercises. The reliability of this benchmark has been confirmed in several ways. Subsequent experiments with existing LLMs have led to the following findings: (i) Chinese medical LLMs underperform on this benchmark, especially where medical reasoning and factual consistency are vital, underscoring the need for advances in clinical knowledge and diagnostic accuracy. (ii) Several general-domain LLMs demonstrate substantial potential in medical clinics, while the limited input capacity of many medical LLMs hinders their practical use. These findings reveal both the strengths and limitations of LLMs in clinical scenarios and offer critical insights for medical research.

Program-of-Thought (PoT) replaces natural language-based Chain-of-Thought (CoT) as the most popular method in Large Language Models (LLMs) mathematical reasoning tasks by utilizing external tool calls to circumvent computational errors. However, our evaluation of the GPT-4 and Llama series reveals that using PoT introduces more reasoning errors, such as incorrect formulas or flawed logic, compared to CoT. To address this issue, we propose Human-Think Language (HTL), which leverages a suite of strategies that help integrate PoT and CoT, encompassing: (1) a new generation paradigm that uses full CoT reasoning to control code generation. (2) Focus Attention, that directs model attention to the CoT reasoning during PoT to generate more logical code. (3) reinforcement learning that utilizes the accuracy of both CoT and PoT responses as rewards to prevent repetitive reasoning steps in LLMs when solving difficult math problems. Our method achieves an average improvement of 6.5% on the Llama-Base model and 4.3% on the Mistral-Base model across 8 mathematical calculation datasets. It also shows significant effectiveness on five out-of-domain datasets by controlling the model’s information flow, exhibiting strong transferability. Additionally, HTL shows the most significant improvement in non-mathematical natural language inference task, contributing to a unified reasoning task framework.

Sentiment analysis is increasingly viewed as a vital task both from an academic and a commercial standpoint. In this paper, we focus on the structured sentiment analysis task that is released on SemEval-2022 Task 10. The task aims to extract the structured sentiment information (e.g., holder, target, expression and sentiment polarity) in a text. We propose a simple and unified model for both the monolingual and crosslingual structured sentiment analysis tasks. We translate this task into an event extraction task by regrading the expression as the trigger word and the other elements as the arguments of the event. Particularly, we first extract the expression by judging its start and end indices. Then, to consider the expression, we design a conditional layer normalization algorithm to extract the holder and target based on the extracted expression. Finally, we infer the sentiment polarity based on the extracted structured information. Pre-trained language models are utilized to obtain the text representation. We conduct the experiments on seven datasets in five languages. It attracted 233 submissions in monolingual subtask and crosslingual subtask from 32 teams. Finally, we obtain the top 5 place on crosslingual tasks.

Generating explanations for recommender systems is essential for improving their transparency, as users often wish to understand the reason for receiving a specified recommendation. Previous methods mainly focus on improving the generation quality, but often produce generic explanations that fail to incorporate user and item specific details. To resolve this problem, we present Multi-Scale Distribution Deep Variational Autoencoders (MVAE).These are deep hierarchical VAEs with a prior network that eliminates noise while retaining meaningful signals in the input, coupled with a recognition network serving as the source of information to guide the learning of the prior network. Further, the Multi-scale distribution Learning Framework (MLF) along with a Target Tracking Kullback-Leibler divergence (TKL) mechanism are proposed to employ multi KL divergences at different scales for more effective learning. Extensive empirical experiments demonstrate that our methods can generate explanations with concrete input-specific contents.