Wenhui Que

Kerui Peng, Feifei Li, Xingyu Fan et al.

When large language models are fine-tuned to generate persona- or tone-conditioned responses, their output diversity is severely limited--a failure we term Cross-Style Collapse. We trace this collapse to the cross-entropy objective, which under shared representations tends to suppress diverse continuations. We propose Semantic Flow Regularization (SFR), a lightweight auxiliary objective that supervises the backbone with continuous sentence-encoder embeddings of future segments via conditional flow matching. The stochastic flow source preserves multi-modality by construction; the flow-matching head is discarded at inference, adding zero deployment cost. On a large-scale industrial dialogue dataset (Qwen3-32B, 9 personas), SFR improves output diversity, style fidelity, and response quality over SFT. We further validate on the public LiveCodeBench-v5 (Qwen2.5-Coder-7B-Instruct), where SFR consistently improves pass@k, confirming generality beyond stylized dialogue. A controlled comparison on MBPP reveals Multi-Token Prediction to be a degenerate special case of SFR.

Zihan Li, Xingyu Fan, Feifei Li et al.

Existing agent memory systems universally follow what we term a Memory-as-Tool paradigm where a single query triggers one-shot retrieval of flat passage lists, suffering from passive invocation, reasoning-retrieval decoupling, and structural mismatch between retrieved fragments and the agent's navigational needs. We propose MemCog, a Memory-as-Cognition system that makes memory access an integral part of the reasoning process. MemCog organizes user knowledge as Navigable Memory Store with associative link graphs, exposes Cross-Dimensional Navigation Interface for multi-step reasoning-driven traversal, and employs Proactive Reasoning Protocol that drives agents to spontaneously initiate memory exploration from conversational context. We additionally construct ProactiveMemBench, the first benchmark for evaluating proactive memory triggering. Experiments show that MemCog achieves state-of-the-art on passive QA benchmarks (92.98 on LoCoMo, 95.8 on LongMemEval) while substantially outperforming baselines on ProactiveMemBench, demonstrating the advantage of Memory-as-Cognition.

Xiaoqing Wu, Feifei Li, Haoliang Ming et al.

Evidence construction systems--chunk retrieval, agent memory, knowledge-graph traversal, and thematic indexing--are evaluated on separate benchmarks with incompatible corpora and metrics, making cross-paradigm diagnosis impossible. We introduce AuthTrace, the first diagnostic benchmark that places all major paradigms on a single corpus and query set by exploiting the dual nature of single-author collections. Built on thematically dense corpora where all texts share style, topic, and vocabulary, AuthTrace provides 2,099 instances with exhaustive gold evidence and a fan-in gradient as the primary diagnostic axis. Comparing eight systems across two QA models, we find that (1) evidence recall--not precision--is the dominant predictor of answer quality (r = 0.96); (2) fan-in exposes paradigm-specific collapse patterns, with flat retrieval degrading 3x faster than structured-evidence systems; and (3) full-context prompting fails uniformly, establishing evidence construction as a necessary capacity beyond raw corpus exposure.

Haoliang Ming, Feifei Li, Xiaoqing Wu et al.

LLM agents require retrieval to behave less like one-shot context fetching and more like reasoning: searching, reading, traversing, and deciding when evidence is sufficient. However, Retrieval-Augmented Generation (RAG) typically organizes external knowledge as flat chunks retrieved by embedding similarity, exposing a retrieval-as-lookup interface that is poorly aligned with tool-using agents. We propose LLM-Wiki, an agent-native retrieval system that operationalizes the Retrieval-as-Reasoning paradigm by treating external knowledge as a compilable, composable, and self-evolving structure rather than a static retrieval index. LLM-Wiki compiles documents into structured Wiki pages with bidirectional links, exposes search, read, and link-following operations through standard tool-calling interfaces, and introduces an Error Book for persistent structural and semantic self-correction. On HotpotQA, MuSiQue, and 2WikiMultiHopQA, LLM-Wiki outperforms seven baselines, including HippoRAG 2, LightRAG, and GraphRAG, with gains of 2.0-8.1 F1 points over the strongest graph-based baseline and larger gains over Dense RAG. On AuthTrace, LLM-Wiki achieves the best overall accuracy, with especially strong gains on multi-document structured queries, showing that compilation-based knowledge organization generalizes beyond chain-style multi-hop reasoning.

Wendong Bi, Yirong Mao, Xianglong Liu et al.

Personalized music recommendation in conversational scenarios usually requires a deep understanding of user preferences and nuanced musical context, yet existing methods often struggle with balancing specialized domain knowledge and flexible tool integration. This paper proposes WeMusic-Agent, a training framework for efficient LLM-based conversational music recommendation. By integrating the knowledge internalization and agentic boundary learning, the framework aims to teach the model to intelligently decide when to leverage internalized knowledge and when to call specialized tools (e.g., music retrieval APIs, music recommendation systems). Under this framework, we present WeMusic-Agent-M1, an agentic model that internalizes extensive musical knowledge via continued pretraining on 50B music-related corpus while acquiring the ability to invoke external tools when necessary. Additionally, considering the lack of open-source benchmarks for conversational music recommendation, we also construct a benchmark for personalized music recommendations derived from real-world data in WeChat Listen. This benchmark enables comprehensive evaluation across multiple dimensions, including relevance, personalization, and diversity of the recommendations. Experiments on real-world data demonstrate that WeMusic-Agent achieves significant improvements over existing models.

Xingyu Fan, Feifei Li, Wenhui Que et al.

Conversational agents deployed in industrial-scale official account platforms must generate responses that are both contextually grounded and stylistically aligned-requirements that existing methods struggle to meet. Chain-of-thought (CoT) prompting induces significant latency due to multi-turn reasoning; per-account fine-tuning is computationally prohibitive; and long prompt-based methods degrade the model's ability to grasp injected context and style. In this paper, we propose WeStar, a lite-adaptive framework for stylized contextual question answering that scales to millions of official accounts. WeStar combines context-grounded generation via RAG with style-aware generation using Parametric RAG (PRAG), where LoRA modules are dynamically activated per style cluster. Our contributions are fourfold: (1) We introduce WeStar, a unified framework capable of serving large volumes of official accounts with minimal overhead. (2) We propose a multi-dimensional, cluster-based parameter sharing scheme that enables compact style representation while preserving stylistic diversity. (3) We develop a style-enhanced Direct Preference Optimization (SeDPO) method to optimize each style cluster's parameters for improved generation quality. (4) Experiments on a large-scale industrial dataset validate the effectiveness and efficiency of WeStar, underscoring its pracitical value in real-world deployment.