Min Zhang
Publications
From Profiling to Synthesis: Benchmarking Implicit Behavioral Alignment in Personalized LLM Agents
Large Language Models have enabled increasingly capable autonomous agents, yet personalization remains critical for making such agents practically useful. Recent benchmarks have begun evaluating personalization in agents, but they largely rely on static preference snapshots, fixed interaction logs, or question answering over predefined user profiles. Such designs fail to capture the complexity of evolving user preferences and neglect preference-conditioned task execution-a discrepancy we term as the knowledge-to-action gap. To address this challenge, we introduce IBA-Bench, a benchmark for implicit behavioral alignment constructed from longitudinal interaction histories that contain noise, implicit cues, and temporal inconsistencies. Unlike prior work, IBA-Bench evaluates whether an agent can execute tasks while satisfying implicit user constraints inferred from historical interactions. We further propose IBA-Agent, an agent framework that reconciles conflicting priorities through broad retrieval and trajectory-level alignment. Experiment results on IBA-Bench show that effective personalization remains a significant challenge for state-of-the-art LLM agents, and the proposed IBA-Agent substantially improves behavioral alignment in complex scenarios across nine application domains.
SLAI T-Rex: Full-Parameter Post-training of the DeepSeek-V4 Family on Ascend SuperPOD
Full-parameter post-training of trillion-parameter-scale MoE models introduces substantial system-level challenges for large-scale distributed training, including severe memory pressure, non-overlapped communication overhead, and inefficient kernel execution. While most large-scale LLM training systems are built around GPU-based clusters, this report presents an end-to-end optimization practice on the Ascend NPU SuperPOD. Using the DeepSeek-V4 model family as the target workload, we develop a hierarchical optimization framework spanning model-level parallelism, computation-communication orchestration, and low-level kernel execution. The resulting system achieves 34.22% Model FLOPs Utilization (MFU) with a 2.93x improvement over the open-source baseline recipe while maintaining training stability. Building on this optimized infrastructure, we further establish a CPT and SFT workflow for complex Operations Research (OR) tasks. We refer to the integrated framework as SLAI T-Rex. Using DeepSeek-V4-Flash, we develop OR-oriented CPT and SFT data pipelines that combine collected domain resources with solver-verified synthetic optimization documents. The resulting dataset contains 10K high-quality SFT samples spanning four task categories and three problem representations. The specialized model achieves the highest average zero-shot Pass@1 score among the evaluated models, reaching 71.81% and outperforming GPT-5.4-Mini and the base DeepSeek-V4-Flash model by 3.98 and 11.27 percentage points, respectively. Overall, this work demonstrates a full-stack pathway from efficient trillion-parameter model post-training on Ascend infra to domain-specialized Flash models for solver-grounded mathematical modeling, advancing frontier-model systems for complex reasoning.
SuCo: Sufficiency-guided Continuous Adaptive Reasoning
Despite remarkable performance on complex tasks, Large Reasoning Models (LRMs) often generate excessively long Chain-of-Thoughts (CoT), inflating computational costs even for simple queries. Existing efforts to mitigate this inefficiency typically rely on discrete reasoning modes or fixed budget tiers, lacking a principled criterion of when reasoning is sufficient. In this work, we introduce Minimal Sufficient CoT (MSC), defined as the shortest prefix of a CoT trajectory which is adequate for producing the correct answer. We empirically show that MSC not only reduces reasoning tokens, but also improves accuracy across difficulty levels. Building on MSC, we propose Sufficiency-guided Continuous Adaptive Reasoning (SuCo), a two-stage training framework for autonomous reasoning control along a continuous spectrum. In stage 1, MSC-Aligned Fine-Tuning (MFT) constructs MSC data using problem-adaptive sufficiency thresholds that naturally scale with question difficulty, then fine-tunes the model to internalize concise yet sufficient reasoning patterns. In stage 2, Sufficiency-Aware Policy Optimization (SAPO) further optimizes the model through reinforcement learning with dynamic complexity tracking and sufficiency-aware rewards that penalize both over- and under-thinking. Extensive experiments across mathematics, code, and science benchmarks show that SuCo consistently achieves improvements in both accuracy and reasoning efficiency.
Taming Actor-Observer Asymmetry in Agents via Dialectical Alignment
Large Language Model agents have rapidly evolved from static text generators into dynamic systems capable of executing complex autonomous workflows. To enhance reliability, multi-agent frameworks assigning specialized roles are increasingly adopted to enable self-reflection and mutual auditing. While such role-playing effectively leverages domain expert knowledge, we find it simultaneously induces a human-like cognitive bias known as Actor-Observer Asymmetry (AOA). Specifically, an agent acting as an actor (during self-reflection) tends to attribute failures to external factors, whereas an observer (during mutual auditing) attributes the same errors to internal faults. We quantify this using our new Ambiguous Failure Benchmark, which reveals that simply swapping perspectives triggers the AOA effect in over 20% of cases for most models. To tame this bias, we introduce ReTAS (Reasoning via Thesis-Antithesis-Synthesis), a model trained through dialectical alignment to enforce perspective-invariant reasoning. By integrating dialectical chain-of-thought with Group Relative Policy Optimization, ReTAS guides agents to synthesize conflicting viewpoints into an objective consensus. Experiments demonstrate that ReTAS effectively mitigates attribution inconsistency and significantly improves fault resolution rates in ambiguous scenarios.
OGER: A Robust Offline-Guided Exploration Reward for Hybrid Reinforcement Learning
Recent advancements in Reinforcement Learning with Verifiable Rewards (RLVR) have significantly improved Large Language Model (LLM) reasoning, yet models often struggle to explore novel trajectories beyond their initial latent space. While offline teacher guidance and entropy-driven strategies have been proposed to address this, they often lack deep integration or are constrained by the model's inherent capacity. In this paper, we propose OGER, a novel framework that unifies offline teacher guidance and online reinforcement learning through a specialized reward modeling lens. OGER employs multi-teacher collaborative training and constructs an auxiliary exploration reward that leverages both offline trajectories and the model's own entropy to incentivize autonomous exploration. Extensive experiments across mathematical and general reasoning benchmarks demonstrate that OGER significantly outperforms competitive baselines, achieving substantial gains in mathematical reasoning while maintaining robust generalization to out-of-domain tasks. We provide a comprehensive analysis of training dynamics and conduct detailed ablation studies to validate the effectiveness of our entropy-aware reward modulation. Our code is available at https://github.com/ecoli-hit/OGER.git.
D-QRELO: Training- and Data-Free Delta Compression for Large Language Models via Quantization and Residual Low-Rank Approximation
Supervised Fine-Tuning (SFT) accelerates taskspecific large language models (LLMs) development, but the resulting proliferation of finetuned models incurs substantial memory overhead. Delta compression addresses this by retaining a single pre-trained LLM with multiple compressed delta weights. However, existing methods fail on models fine-tuned with largescale datasets. We find that larger SFT data scale amplifies delta parameter magnitude, singular values, and entropy, exacerbating compression errors. To tackle this, we propose DQRELO (Delta Compression via Quantization and Residual Low-Rank), a novel training- and data-free delta compression method. It combines coarse-grained one-bit quantization to capture the dominant structure of the delta, followed by compensated residual low-rank approximation to recover fine-grained details from the smaller residual error. Experiments on various LLMs spanning dense and MoE architectures across multiple domains under this challenging setting demonstrate that DQRELO outperforms existing methods. Moreover, we establish key design principles for delta compression through extensive empirical analysis, demonstrating how task difficulty, architecture, and layer positioning create predictable patterns that can guide optimal compression strategies in production systems.
Step-GRPO: Internalizing Dynamic Early Exit for Efficient Reasoning
Large reasoning models that use long chain-of-thought excel at problem-solving yet waste compute on redundant checks. Curbing this overthinking is hard: training-time length penalties can cripple ability, while inference-time early-exit adds system overhead. To bridge this gap, we propose Step-GRPO, a novel post-training framework that internalizes dynamic early-exit capabilities directly into the model. Step-GRPO shifts the optimization objective from raw tokens to semantic steps by utilizing linguistic markers to structure reasoning. We introduce a Dynamic Truncated Rollout mechanism that exposes the model to concise high-confidence trajectories during exploration, synergized with a Step-Aware Relative Reward that dynamically penalizes redundancy based on group-level baselines. Extensive experiments across three model sizes on diverse benchmarks demonstrate that Step-GRPO achieves a superior accuracy-efficiency trade-off. On Qwen3-8B, our method reduces token consumption by 32.0\% compared to the vanilla model while avoiding the accuracy degradation observed in traditional length-penalty methods.
MathAgent: Adversarial Evolution of Constraint Graphs for Mathematical Reasoning Data Synthesis
Synthesizing high-quality mathematical reasoning data without human priors remains a significant challenge. Current approaches typically rely on seed data mutation or simple prompt engineering, often suffering from mode collapse and limited logical complexity. This paper proposes a hierarchical synthesis framework that formulates data synthesis as an unsupervised optimization problem over a constraint graph followed by semantic instantiation, rather than treating it as a direct text generation task. We introduce a Legislator-Executor paradigm: The Legislator adversarially evolves structured generation blueprints encoding the constraints of the problem, while the Executor instantiates these specifications into diverse natural language scenarios. This decoupling of skeleton design from linguistic realization enables a prioritized focus on constructing complex and diverse logical structures, thereby guiding high-quality data synthesis. Experiments conducted on a total of 10 models across the Qwen, Llama, Mistral, and Gemma series demonstrate that our method achieves notable results: models fine-tuned on 1K synthesized samples outperform widely-used datasets of comparable scale (LIMO, s1K) across eight mathematical benchmarks, exhibiting superior out-of-distribution generalization.
E3-TIR: Enhanced Experience Exploitation for Tool-Integrated Reasoning
While Large Language Models (LLMs) have demonstrated significant potential in Tool-Integrated Reasoning (TIR), existing training paradigms face significant limitations: Zero-RL suffers from inefficient exploration and mode degradation due to a lack of prior guidance, while SFT-then-RL is limited by high data costs and capability plateaus caused by low-entropy collapse. To address these challenges, we propose E3-TIR (Enhanced Experience Exploitation), a warm-up paradigm for the early stages of agent training. Specifically, we formulate training as the dynamic integration of three experience types: Expert Prefixes, Expert Guided, and Self-Exploration. By executing diverse branching exploration around expert "anchors" and employing a mix policy optimization mechanism, we effectively mitigate distribution shifts and resolve optimization conflicts arising from shared prefixes. Our method dynamically adapts the model's knowledge boundaries, effectively balancing exploration diversity with training efficiency.Experimental results demonstrate that E3-TIR achieves a 6 performance improvement over traditional paradigms on tool-use tasks, while requiring less than 10 of the synthetic data. Furthermore, in terms of ROI, a comprehensive metric integrating performance, data cost, and training efficiency we achieve a 1.46x gain compared to baselines. Code is available at https://github.com/yuki-younai/E3-TIR.
Backdoors in RLVR: Jailbreak Backdoors in LLMs From Verifiable Reward
Reinforcement Learning with Verifiable Rewards (RLVR) is an emerging paradigm that significantly boosts a Large Language Model's (LLM's) reasoning abilities on complex logical tasks, such as mathematics and programming. However, we identify, for the first time, a latent vulnerability to backdoor attacks within the RLVR framework. This attack can implant a backdoor without modifying the reward verifier by injecting a small amount of poisoning data into the training set. Specifically, we propose a novel trigger mechanism designated as the \ourapproach (ACB). The attack exploits the RLVR training loop by assigning substantial positive rewards for harmful responses and negative rewards for refusals. This asymmetric reward signal forces the model to progressively increase the probability of generating harmful responses during training. Our findings demonstrate that the RLVR backdoor attack is characterized by both high efficiency and strong generalization capabilities. Utilizing less than 2\% poisoned data in train set, the backdoor can be successfully implanted across various model scales without degrading performance on benign tasks. Evaluations across multiple jailbreak benchmarks indicate that activating the trigger degrades safety performance by an average of 73\%. Furthermore, the attack generalizes effectively to a wide range of jailbreak methods and unsafe behaviors. Code is available at https://github.com/yuki-younai/Backdoor_in_RLVR.