arXiv subjects

Bolin Ding

Publications and source records attributed to Bolin Ding.

At least 37 records · Page 2Linked to original sources

arXiv2026

Retrieval Heads are Dynamic

Recent studies have identified "retrieval heads" in Large Language Models (LLMs) responsible for extracting information from input contexts. However, prior works largely rely on static statistics aggregated across datasets, identifying heads that perform retrieval on average. This perspective overlooks the fine-grained temporal dynamics of autoregressive generation. In this paper, we investigate retrieval heads from a dynamic perspective. Through extensive analysis, we establish three core claims: (1) Dynamism: Retrieval heads vary dynamically across timesteps; (2) Irreplaceability: Dynamic retrieval heads are specific at each timestep and cannot be effectively replaced by static retrieval heads; and (3) Correlation: The model's hidden state encodes a predictive signal for future retrieval head patterns, indicating an internal planning mechanism. We validate these findings on the Needle-in-a-Haystack task and a multi-hop QA task, and quantify the differences on the utility of dynamic and static retrieval heads in a Dynamic Retrieval-Augmented Generation framework. Our study provides new insights into the internal mechanisms of LLMs.

cs.CL↗

arXiv2026

Large Language Model-Enhanced Relational Operators: Taxonomy, Benchmark, and Analysis

With the development of large language models (LLMs), numerous studies integrate LLMs through operator-like components to enhance relational data processing tasks, e.g., filters with semantic predicates, knowledge-augmented table imputation, reasoning-driven entity matching and more challenging semantic query processing. These components invoke LLMs while preserving a relational input/output interface, which we refer to as LLM-Enhanced Relational Operators (LROs). From an operator perspective, unfortunately, these existing LROs suffer from fragmented definition, various implementation strategies and inadequate evaluation benchmarks. To this end, in this paper, we first establish a unified LRO taxonomy to align existing LROs, and categorize them into: Select, Match, Impute, Cluster and Order, along with their operands and implementation variants. Second, we design LROBench, a comprehensive benchmark featuring 290 single-LRO queries and 60 multi-LRO queries, spanning 27 databases across more than 10 domains. LROBench covers all operating logics and operand granularities in its single-LRO workload, and provides challenging multi-LRO queries stratified by query complexity. Based on these, we evaluate individual LROs under various implementations, deriving practical insights into LRO design choices and summarizing our empirical best practices. We further compare the end-to-end performance of existing multi-LRO systems against an LRO suite instantiated with these best practices, in order to investigate how to design an effective LRO set for multi-LRO systems targeting complex semantic queries. Last, to facilitate future work, we outline promising future directions and open-source all benchmark data and evaluation code, available at https://github.com/LROBench/LROBench/.

cs.DB↗

arXiv2026

Remember Me, Refine Me: A Dynamic Procedural Memory Framework for Experience-Driven Agent Evolution

Procedural memory enables large language model (LLM) agents to internalize "how-to" knowledge, theoretically reducing redundant trial-and-error. However, existing frameworks predominantly suffer from a "passive accumulation" paradigm, treating memory as a static append-only archive. To bridge the gap between static storage and dynamic reasoning, we propose $\textbf{ReMe}$ ($\textit{Remember Me, Refine Me}$), a comprehensive framework for experience-driven agent evolution. ReMe innovates across the memory lifecycle via three mechanisms: 1) $\textit{multi-faceted distillation}$, which extracts fine-grained experiences by recognizing success patterns, analyzing failure triggers and generating comparative insights; 2) $\textit{context-adaptive reuse}$, which tailors historical insights to new contexts via scenario-aware indexing; and 3) $\textit{utility-based refinement}$, which autonomously adds valid memories and prunes outdated ones to maintain a compact, high-quality experience pool. Extensive experiments on BFCL-V3 and AppWorld demonstrate that ReMe establishes a new state-of-the-art in agent memory system. Crucially, we observe a significant memory-scaling effect: Qwen3-8B equipped with ReMe outperforms larger, memoryless Qwen3-14B, suggesting that self-evolving memory provides a computation-efficient pathway for lifelong learning. We release our code and the $\texttt{reme.library}$ dataset to facilitate further research.

cs.AI↗

arXiv2026

Automated Profile Inference with Language Model Agents

Impressive progress has been made in automated problem-solving by the collaboration of large language model (LLM) based agents. However, these automated capabilities also open avenues for malicious applications. In this paper, we study a new threat that LLMs pose to online pseudonymity, called automated profile inference, where an adversary can instruct LLMs to automatically collect and extract sensitive personal attributes from publicly available user activities on pseudonymous platforms. We also introduce an automated profiling framework called AutoProfiler to demonstrate and assess the feasibility of such attacks in real-world scenarios. AutoProfiler consists of four specialized LLM agents that work collaboratively to retrieve and process user online activities and generate a profile with extracted personal information. Experimental results on two real-world datasets and one synthetic dataset show that AutoProfiler is highly effective and efficient, and the inferred attributes are both identifiable and sensitive, posing significant privacy risks. We explore mitigation strategies from different perspectives and advocate for increased public awareness of this emerging privacy threat.

cs.CR↗

arXiv2026

HumanVBench: Probing Human-Centric Video Understanding in MLLMs with Automatically Synthesized Benchmarks

Evaluating the nuanced human-centric video understanding capabilities of Multimodal Large Language Models (MLLMs) remains a great challenge, as existing benchmarks often overlook the intricacies of emotion, behavior, and cross-modal alignment. We introduce HumanVBench, a comprehensive video benchmark designed to rigorously probe these capabilities across 16 fine-grained tasks. A cornerstone of our work is a novel and scalable benchmark construction methodology, featuring two automated pipelines that synthesize high-quality video annotations and challenging multiple-choice questions with minimal human labor. By leveraging state-of-the-art models for annotation and systematically converting model-induced errors into plausible distractors, our framework provides a generalizable ``machine'' for creating nuanced evaluation suites. Our extensive evaluation of 30 leading MLLMs on HumanVBench reveals critical deficiencies, particularly in perceiving subtle emotions and aligning speech with visual cues, with even top proprietary models falling short of human performance. We open-source HumanVBench and our synthesis pipelines to catalyze the development of more socially intelligent and capable video MLLMs.

cs.CV↗

arXiv2026

E2E-REME: Towards End-to-End Microservices Auto-Remediation via Experience-Simulation Reinforcement Fine-Tuning

Contemporary microservice systems continue to grow in scale and complexity, leading to increasingly frequent and costly failures. While recent LLM-based auto-remediation approaches have emerged, they primarily translate textual instructions into executable Ansible playbooks and rely on expert-crafted prompts, lacking runtime knowledge guidance and depending on large-scale general-purpose LLMs, which limits their accuracy and efficiency. We introduce \textit{End-to-End Microservice Remediation} (E2E-MR), a new task that requires directly generating executable playbooks from diagnosis reports to autonomously restore faulty systems. To enable rigorous evaluation, we build \textit{MicroRemed}, a benchmark that automates microservice deployment, failure injection, playbook execution, and post-repair verification. We further propose \textit{E2E-REME}, an end-to-end auto-remediation model trained via experience-simulation reinforcement fine-tuning. Experiments on public and industrial microservice platforms, compared with nine representative LLMs, show that E2E-REME achieves superior accuracy and efficiency.

cs.SE↗

arXiv2026

Talk to Right Specialists: Iterative Routing in Multi-agent Systems for Question Answering

Retrieval-augmented generation (RAG) agents are increasingly deployed to answer questions over local knowledge bases that cannot be centralized due to knowledge-sovereignty constraints. This results in two recurring failures in production: users do not know which agent to consult, and complex questions require evidence distributed across multiple agents. To overcome these challenges, we propose RIRS, a training-free orchestration framework to enable a multi-agent system for question answering. In detail, RIRS summarizes each agent's local corpus in an embedding space, enabling a user-facing server to route queries only to the most relevant agents, reducing latency and avoiding noisy "broadcast-to-all" contexts. For complicated questions, the server can iteratively aggregate responses to derive intermediate results and refine the question to bridge the gap toward a comprehensive answer. Extensive experiments demonstrate the effectiveness of RIRS, including its ability to precisely select agents and provide accurate responses to single-hop queries, and its use of an iterative strategy to achieve accurate, multi-step resolutions for complex queries.

cs.MA↗

arXiv2026

XiYan-SQL: A Novel Multi-Generator Framework For Text-to-SQL

To leverage the advantages of LLM in addressing challenges in the Text-to-SQL task, we present XiYan-SQL, an innovative framework effectively generating and utilizing multiple SQL candidates. It consists of three components: 1) a Schema Filter module filtering and obtaining multiple relevant schemas; 2) a multi-generator ensemble approach generating multiple highquality and diverse SQL queries; 3) a selection model with a candidate reorganization strategy implemented to obtain the optimal SQL query. Specifically, for the multi-generator ensemble, we employ a multi-task fine-tuning strategy to enhance the capabilities of SQL generation models for the intrinsic alignment between SQL and text, and construct multiple generation models with distinct generation styles by fine-tuning across different SQL formats. The experimental results and comprehensive analysis demonstrate the effectiveness and robustness of our framework. Overall, XiYan-SQL achieves a new SOTA performance of 75.63% on the notable BIRD benchmark, surpassing all previous methods. It also attains SOTA performance on the Spider test set with an accuracy of 89.65%.

cs.CL↗

arXiv2026

FIPO: Eliciting Deep Reasoning with Future-KL Influenced Policy Optimization

We present Future-KL Influenced Policy Optimization (FIPO), a reinforcement learning algorithm designed to overcome reasoning bottlenecks in large language models. While GRPO style training scales effectively, it typically relies on outcome-based rewards (ORM) that distribute a global advantage uniformly across every token in a trajectory. We argue that this coarse-grained credit assignment imposes a performance ceiling by failing to distinguish critical logical pivots from trivial tokens. FIPO addresses this by incorporating discounted future-KL divergence into the policy update, creating a dense advantage formulation that re-weights tokens based on their influence on subsequent trajectory behavior. Empirically, FIPO enables models to break through the length stagnation seen in standard baselines. Evaluated on Qwen2.5-32B, FIPO extends the average chain-of-thought length from roughly 4,000 to over 10,000 tokens and increases AIME 2024 Pass@1 accuracy from 50.0% to a peak of 58.0% (converging at approximately 56.0\%). This outperforms both DeepSeek-R1-Zero-Math-32B (around 47.0%) and o1-mini (approximately 56.0%). Our results suggest that establishing dense advantage formulations is a vital path for evolving ORM-based algorithms to unlock the full reasoning potential of base models. We open-source our training system, built on the verl framework.

cs.LG↗

arXiv2026

AmbiSQL: Interactive Ambiguity Detection and Resolution for Text-to-SQL

Text-to-SQL systems translate natural language questions into SQL queries, providing substantial value for non-expert users. While large language models (LLMs) show promising results for this task, they remain error-prone. Query ambiguity has been recognized as a major obstacle in LLM-based Text-to-SQL systems, leading to misinterpretation of user intent and inaccurate SQL generation. To this end, we present AmbiSQL, an interactive system that automatically detects query ambiguities and guides users through intuitive multiple-choice questions to clarify their intent. It introduces a fine-grained ambiguity taxonomy for identifying ambiguities arising from both database elements and LLM reasoning, and subsequently incorporates user feedback to rewrite ambiguous questions. In this demonstration, AmbiSQL is integrated with XiYan-SQL, our commercial Text-to-SQL backend. We provide 40 ambiguous queries collected from two real-world benchmarks that SIGMOD'26 attendees can use to explore how disambiguation improves SQL generation quality. Participants can also apply the system to their own databases and natural language questions. The codebase and demo video are available at: https://github.com/JustinzjDing/AmbiSQL and https://www.youtube.com/watch?v=rbB-0ZKwYkk.

cs.DB↗

arXiv2026

On the Direction of RLVR Updates for LLM Reasoning: Identification and Exploitation

Reinforcement learning with verifiable rewards (RLVR) has substantially improved the reasoning capabilities of large language models. While existing analyses identify that RLVR-induced changes are sparse, they primarily focus on the \textbf{magnitude} of these updates, largely overlooking their \textbf{direction}. In this work, we argue that the direction of updates is a more critical lens for understanding RLVR's effects, which can be captured by the signed, token-level log probability difference $Δ\log p$ between the base and final RLVR models. Through statistical analysis and token-replacement interventions, we demonstrate that $Δ\log p$ more effectively identifies sparse, yet reasoning-critical updates than magnitude-based metrics (\eg divergence or entropy). Building on this insight, we propose two practical applications: (1) a \textit{test-time extrapolation} method that amplifies the policy along the learned $Δ\log p$ direction to improve reasoning accuracy without further training; (2) a \textit{training-time reweighting} method that focuses learning on low-probability (corresponding to higher $Δ\log p$) tokens, which improves reasoning performance across models and benchmarks. Our work establishes the direction of change as a key principle for analyzing and improving RLVR.

cs.LG↗

arXiv2026

Sparse but Critical: A Token-Level Analysis of Distributional Shifts in RLVR Fine-Tuning of LLMs

Reinforcement learning with verifiable rewards (RLVR) has significantly improved reasoning in large language models (LLMs), yet the token-level mechanisms underlying these improvements remain unclear. We present a systematic empirical study of RLVR's distributional effects organized around three main analyses: (1) token-level characterization of distributional shifts between base and RL models, (2) the impact of token-level distributional shifts on sequence-level reasoning performance through cross-sampling interventions, and (3) fine-grained mechanics of these shifts at the token level. We find that RL fine-tuning induces highly sparse and targeted changes, with only a small fraction of token distributions exhibiting meaningful divergence between the base and RL policies. We further characterize the structure and evolution of these shifts through analyses of token entropy, positional concentration, and reallocation of probability mass. To assess the functional importance of these sparse changes, we conduct cross-sampling experiments that selectively swap token choices between the base and RL models with varying intervention budgets. We show that inserting only a small fraction of RL-sampled tokens into base generations progressively recovers RL performance gains, while injecting a similarly small number of base token choices into otherwise RL-generated sequences collapses performance to base levels, isolating a small set of token-level decisions directly responsible for RLVR's performance gains. Finally, we explore divergence-weighted variants of the advantage signal as a diagnostic intervention, finding that they can yield improvements over baselines. Together, our results shed light on the distributional changes induced by RLVR and provide a fine-grained, token-level lens for understanding RLVR fine-tuning as a targeted refinement process.

cs.CL↗

arXiv2026

On-Policy RL Meets Off-Policy Experts: Harmonizing Supervised Fine-Tuning and Reinforcement Learning via Dynamic Weighting

Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) are two prominent post-training paradigms for refining the capabilities and aligning the behavior of Large Language Models (LLMs). Existing approaches that integrate SFT and RL often face the risk of disrupting established response patterns and inducing overfitting to expert data. To address this, we present a novel investigation into the unified view of SFT and RL through an off-policy versus on-policy lens. We propose CHORD, a framework for Controllable Harmonization of On- and Off-Policy Reinforcement Learning via Dynamic Weighting, which reframes SFT not as a separate stage but as a dynamically weighted auxiliary objective within the on-policy RL process. Based on an analysis of off-policy expert data's influence at both holistic and granular levels, we incorporate a dual-control mechanism in CHORD. Specifically, the framework first employs a global coefficient to holistically guide the transition from off-policy imitation to on-policy exploration, and then applies a token-wise weighting function that enables granular learning from the expert, which promotes on-policy exploration and mitigates disruption from off-policy data. We conduct extensive experiments across various practical tasks, providing empirical evidence that CHORD achieves a stable and efficient learning process. By effectively harmonizing off-policy expert data with on-policy exploration, CHORD demonstrates significant improvements over baselines. We release the implementation at https://github.com/modelscope/Trinity-RFT/tree/main/examples/mix_chord to inspire further research.

cs.LG↗

arXiv2026

Incentivizing Strong Reasoning from Weak Supervision

Large language models (LLMs) have demonstrated impressive performance on reasoning-intensive tasks, but enhancing their reasoning abilities typically relies on either reinforcement learning (RL) with verifiable signals or supervised fine-tuning (SFT) with high-quality long chain-of-thought (CoT) demonstrations, both of which are expensive. In this paper, we study a novel problem of incentivizing the reasoning capacity of LLMs without expensive high-quality demonstrations and reinforcement learning. We investigate whether the reasoning capabilities of LLMs can be effectively incentivized via supervision from significantly weaker models. We further analyze when and why such weak supervision succeeds in eliciting reasoning abilities in stronger models. Our findings show that supervision from significantly weaker reasoners can substantially improve student reasoning performance, recovering close to 94% of the gains of expensive RL at a fraction of the cost. Experiments across diverse benchmarks and model architectures demonstrate that weak reasoners can effectively incentivize reasoning in stronger student models, consistently improving performance across a wide range of reasoning tasks. Our results suggest that this simple weak-to-strong paradigm is a promising and generalizable alternative to costly methods for incentivizing strong reasoning capabilities at inference-time in LLMs. The code is publicly available at https://github.com/yuanyige/w2sr.

cs.CL↗

arXiv2026

Group-Relative REINFORCE Is Secretly an Off-Policy Algorithm: Demystifying Some Myths About GRPO and Its Friends

Off-policy reinforcement learning (RL) for large language models (LLMs) is attracting growing interest, driven by practical constraints in real-world applications, the complexity of LLM-RL infrastructure, and the need for further innovations of RL methodologies. While classic REINFORCE and its modern variants like Group Relative Policy Optimization (GRPO) are typically regarded as on-policy algorithms with limited tolerance of off-policyness, we present in this work a first-principles derivation for group-relative REINFORCE -- a REINFORCE variant that uses the within-group mean reward as the baseline for advantage calculation -- without assuming a specific training data distribution, showing that it admits a native off-policy interpretation. This perspective yields two general principles for adapting REINFORCE to truly off-policy settings: regularizing policy updates, and actively shaping the data distribution. Our analysis demystifies some myths about the roles of importance sampling and clipping in GRPO, unifies and reinterprets two recent algorithms -- Online Policy Mirror Descent and Asymmetric REINFORCE -- as regularized forms of the REINFORCE loss, and offers theoretical justification for seemingly heuristic data-weighting strategies. Our findings lead to actionable insights that are validated with extensive empirical studies, and open up new opportunities for principled algorithm design in off-policy RL for LLMs. Source code for this work is available at https://github.com/agentscope-ai/Trinity-RFT/tree/main/examples/rec_gsm8k.

cs.LG↗

arXiv2026

Learned Query Optimizer in Alibaba MaxCompute: Challenges, Analysis, and Solutions

Existing learned query optimizers remain ill-suited to modern distributed, multi-tenant data warehouses due to idealized modeling assumptions and design choices. Using Alibaba's MaxCompute as a representative, we surface four fundamental, system-agnostic challenges for any deployable learned query optimizer: 1) highly dynamic execution environments that induce large variance in plan costs; 2) potential absence of input statistics needed for cost estimation; 3) infeasibility of conventional model refinement; and 4) uncertain benefits across different workloads. These challenges expose a deep mismatch between theoretical advances and production realities and demand a principled, deployment-first redesign of learned optimizers. To bridge this gap, we present LOAM, a one-stop learned query optimization framework for MaxCompute. Its design principles and techniques generalize and are readily adaptable to similar systems. Architecturally, LOAM introduces a statistics-free plan encoding that leverages operator semantics and historical executions to infer details about data distributions and explicitly encodes the execution environments of training queries to learn their impacts on plan costs. For online queries with unknown environments at prediction time, LOAM provides a theoretical bound on the achievable performance and a practical strategy to smooth the environmental impacts on cost estimations. For system operating, LOAM integrates domain adaptation techniques into training to generalize effectively to online query plans without requiring conventional refinement. Additionally, LOAM includes a lightweight project selector to prioritize high-benefit deployment projects. LOAM has seen up to 30% CPU cost savings over MaxCompute's native query optimizer on production workloads, which could translate to substantial real-world resource savings.

cs.DB↗

arXiv2026

SeeUPO: Sequence-Level Agentic-RL with Convergence Guarantees

Reinforcement learning (RL) has emerged as the predominant paradigm for training large language model (LLM)-based AI agents. However, existing backbone RL algorithms lack verified convergence guarantees in agentic scenarios, especially in multi-turn settings, which can lead to training instability and failure to converge to optimal policies. In this paper, we systematically analyze how different combinations of policy update mechanisms and advantage estimation methods affect convergence properties in single/multi-turn scenarios. We find that REINFORCE with Group Relative Advantage Estimation (GRAE) can converge to the globally optimal under undiscounted conditions, but the combination of PPO & GRAE breaks PPO's original monotonic improvement property. Furthermore, we demonstrate that mainstream backbone RL algorithms cannot simultaneously achieve both critic-free and convergence guarantees in multi-turn scenarios. To address this, we propose SeeUPO (Sequence-level Sequential Update Policy Optimization), a critic-free approach with convergence guarantees for multi-turn interactions. SeeUPO models multi-turn interaction as sequentially executed multi-agent bandit problems. Through turn-by-turn sequential policy updates in reverse execution order, it ensures monotonic improvement and convergence to global optimal solution via backward induction. Experiments on AppWorld and BFCL v4 demonstrate SeeUPO's substantial improvements over existing backbone algorithms: relative gains of 43.3%-54.6% on Qwen3-14B and 24.1%-41.9% on Qwen2.5-14B (averaged across benchmarks), along with superior training stability.

cs.AI↗

arXiv2026

Auto-Rubric: Learning From Implicit Weights to Explicit Rubrics for Reward Modeling

Conventional reward modeling relies on gradient descent over neural weights, creating opaque, data-hungry "black boxes." We propose a paradigm shift from implicit to explicit reward parameterization, recasting optimization from continuous weight spaces to the discrete space of natural language rubrics. We introduce a training-free framework based on iterative rubric learning: it locally induces discriminative criteria via verification-driven refinement, and globally compresses the candidate criteria pool into a compact core set by maximizing an information-theoretic coding rate objective. We organize the compressed core set into a hierarchical rubric structure -- high-level evaluation dimensions supported by concrete verification checks -- serving as an interpretable, portable reward function. Empirically, our approach challenges prevailing data scaling assumptions: using only 70 preference pairs, our rubric-guided judges outperform fully trained reward models on diverse benchmarks. For instance, Qwen3-8B equipped with our learned rubrics achieves 80.91% on RewardBench2, surpassing the specialized Skywork-Reward-V2-Qwen3-8B (78.20%). These results demonstrate that alignment signals are highly compressible and can be effectively captured through explicit symbolic search.

cs.LG↗