Searcharxiv⌕ Search

arXiv subjects

Daoan Zhang

Publications and source records attributed to Daoan Zhang.

At least 19 recordsLinked to original sources

Veritas: A Verifier Generation Harness for Agent Tasks

The effectiveness of a self-improving system relies heavily on the quality of its verifier, which judges the outcome of each improvement round. However, the verifiers of most tasks are written by hand, which is expensive and does not scale to tasks that come without one. We propose Veritas, a harness that generates verifiers automatically. Unlike general-purpose agent harnesses, Veritas adds three mechanisms for verifier generation: an evidence header stating what the verifier will observe, a library of domain rules that tasks leave unstated, and authoring principles, partly refined by a self-improvement loop. We evaluate on six benchmarks: four scored by execution, whose official verifiers we first certify, and two that compare pairs of responses. With the model fixed, the verifiers Veritas generates agree better with the official verdicts than those of seven other harnesses on all four execution-scored benchmarks. On the two pairwise benchmarks they are at or near the best. They also agree better than those of the vendor's own harness for two of three other model families. Its mechanisms also improve 9 of 12 harness-benchmark pairs scored by execution across three harnesses.

cs.AI↗

MaPPO: Maximum a Posteriori Preference Optimization with Prior Knowledge

As the era of large language models (LLMs) unfolds, Preference Optimization (PO) methods have become a central approach to aligning LLMs with human preferences and improving performance. We propose Maximum a Posteriori Preference Optimization (MaPPO), a methodology for learning from preferences that explicitly incorporates prior reward knowledge into the optimization objective. Building on the paradigm employed by Direct Preference Optimization (DPO) and its variants of treating preference learning as a Maximum Likelihood Estimation (MLE) problem, MaPPO integrates prior reward estimates into a principled Maximum a Posteriori (MaP) objective. This not only generalizes DPO and its variants, but also enhances alignment by mitigating the oversimplified binary classification of responses. Additionally, MaPPO introduces no additional hyperparameters, and supports preference optimization in both offline and online settings. In addition, MaPPO can be used as a plugin for DPO variants, including widely used SimPO, IPO, and CPO, and produce consistent improvements. Extensive empirical evaluations of different model sizes and model series on three standard benchmarks (MT-Bench, AlpacaEval 2.0, and Arena-Hard) demonstrate consistent improvements in alignment performance without sacrificing computational efficiency.

cs.LG↗

RLiD: Reinforcement Learning in the Decision Plane for Language Agents and LLMs

Reinforcement learning (RL) has improved the capabilities of large language models (LLMs) across a wide range of tasks. However, token-level RL assigns an action's advantage to the entire generated response, including tokens specifying the parsed action, reasoning chains, and malformed content. Although the environment executes only the parsed action, variation in how the action is expressed in natural language can introduce additional variance into policy-gradient estimates. We propose RLiD (Reinforcement Learning in the Decision plane), a framework that directly optimizes the probabilities of executable action options. RLiD compiles observations, an optional compact memory, and a labeled action menu into the model input, and uses a large decision model (LDM) to select an action in a single forward pass. By eliminating autoregressive token generation and bypassing action parsing, RLiD avoids invalid action outputs and improves RL rollout efficiency. Our analysis shows that, within a fixed text policy, marginalizing over alternative realizations of the same action removes a zero-mean gradient component without changing the expected update. We further distill trained LDM policies into generative LLMs through supervision on their action trajectories. Experiments on ARC, Countdown, and Sokoban demonstrate improved performance or lower computational cost across several settings. On ALFWorld, the complete training and distillation pipeline reaches 99.3% success in 12.7 H100 GPU-hours, compared with 86.6% in 17.3 GPU-hours for token-level RL of the same LLM. With compiled memory, our strongest Qwen2.5-1.5B LDM and distilled generative LLM achieve 100% success on all 134 unseen and 140 seen ALFWorld evaluation games.

cs.AI↗

VerifierBench: Benchmarking the Generation of Executable Verifiers for LLM Agents

Agents need reliable ways to validate their outputs, both when completing tasks and when filtering synthetic training data. We study verifier generation: from the instruction and environment available for solving a task, an agent writes an executable verifier that must accept every valid solution and reject invalid ones. We introduce VerifierBench, a benchmark of 222 tasks from six agent benchmarks spanning tool-call, state-based, and code verification, with 4,993 hidden test cases that include alternative valid solutions and attacks; Verifier4Verifier (V4V), a six-check audit, screens its source verifiers and test suites. Across 19 models in a common agent harness, normalized balanced accuracy ranges from 10.6 to 62.1, with generated verifiers consistently rejecting invalid solutions more reliably than they accept valid ones. This apparent strength can mask overly restrictive verification: most false rejections in the analyzed records come from verifiers that accept no tested solution to a task. Case studies reveal unsupported assumptions about solution formats, tool use, and code changes that exclude valid solutions. These errors survive self-testing because agents test solutions that share the verifier's mistaken assumptions, or submit verifiers without confirming that any correct solution passes. These findings suggest testing both whether a verifier accepts a correct solution and whether it accepts other solutions permitted by the task. The project page and dataset are publicly available.

cs.SE↗

Many Steps into One: Reasoning Contraction and the Limits of Safe Memory Reuse for LLM Agents

Thought templates, skill libraries, agent memories and caches turn many reasoning steps into one. We call this reasoning contraction. A contracted memory saves reasoning steps and errors until its facts change, then quietly answers wrong. Most systems trust memories by counting successes and failures. We prove that such counters are badly tuned change detectors: a symmetric counter notices a partial change only after a power of its false-alarm horizon, and an unclipped one may never notice it. On five LLMs answering multi-hop questions over edited facts, ExpeL- and ACE-style counters serve 7.0-13.5x more stale answers than certified memories given the same information, and on code tasks with an exact verifier 35-73x more. Safety also has a price. When facts change and each memory is judged only by its own verified outcomes, one number, the certification load, sets how much reasoning any safe memory can skip, and under Zipf demand it yields a scaling law. Our recipe weighs each memory's failures by its own likelihood ratio, certifies its utility against live reasoning, and restarts the certificate when facts change, with the likelihood-ratio detector as the safety net for unannounced changes; for large edits it also invalidates the memory.

cs.CL↗

Large Decision Models: Repurposing Large Language Models as Single-Pass Decision Makers

Large language models make decisions through parallel prefill and sequential decoding. Although effective, this approach incurs decoding costs and does not explicitly expose decision probabilities. Existing task-specific classifiers require per-task training and offer limited generalization to unseen decision schemas. We propose Large Decision Models (LDMs), repurposing language models to return explicit probabilities over labels specified at run time using one prefill forward pass. We first adapt models to this interface through Diverse Decision Supervision (DDS), covering varied domains, rules and labels. To further narrow the performance gap to a reasoning model, Self-Posed Reasoning Distillation (SPRD) adds tasks generated by a frozen teacher, filters them for answer consistency, and trains on label distributions obtained from repeated teacher reasoning. We evaluate on DecisionVista, pairing new-schema tests with held-out human-labelled sources. Across three base models, the recipe improves accuracy over a step-matched baseline by 1.8 and 1.5 percentage points, respectively, achieving a 182-fold speedup in latency over reasoning rollout on Gemma-4-31B. Reasoning remains stronger on the new-schema test, motivating a confidence router for uncertain decisions. On development sets, routing recovers a median 92% of reasoning's accuracy advantage while generating 27-58% of its tokens. Code is available at https://github.com/JianhuiWei7/LDM.

cs.AI↗

Words to World: Benchmarking LLMs in Interactive Physical Environments

Words are increasingly becoming an interface for action: large language models (LLMs) can now invoke tools, operate devices, and make decisions that affect the physical world. However, current benchmarks largely evaluate LLMs through static text responses, offline tool calls, or simulated tasks, leaving open the question of how well they can act in real physical environments. We present a benchmark for measuring LLMs in closed-loop physical interaction, where models control a harness system connected to real-world components including robotic arms, cameras, and sensors. In each task, the model must issue instructions, receive physical feedback from the environment, interpret the resulting state, and adapt its next action accordingly. This setting evaluates capabilities beyond language understanding, including grounded perception, action planning, tool selection, spatial reasoning, feedback-driven correction, and robustness to execution uncertainty. Through experiments with representative LLMs, we characterize their ability to translate linguistic reasoning into physical actions and identify common failure modes such as hallucinated state changes, poor recovery from unexpected outcomes, and unstable long-horizon plans. Our benchmark aims to provide a systematic foundation for evaluating LLMs as interactive agents that connect words, tools, and the physical world.

cs.AI↗

WorldGenBench: A World-Knowledge-Integrated Benchmark for Reasoning-Driven Text-to-Image Generation

Recent text-to-image (T2I) models achieve high visual fidelity and accurate explicit grounding from input prompts. However, image generation is fundamentally a process of semantic expansion. Beyond surface-level descriptions, generating authentic visual content requires supplementing prompts with unstated cues, a crucial step that current models frequently overlook. In response, we introduce WorldGenBench, a benchmark designed to evaluate how well T2I systems integrate factual grounding with contextual inference. Unlike prior work that isolates these abilities, WorldGenBench emphasizes their synergy, recognizing that true understanding requires both knowing and reasoning. To enable interpretable evaluation, we propose the Knowledge Checklist Score, which assesses whether generated images faithfully reflect key explicit and implicit semantic elements. Evaluating 24 models reveals persistent gaps: proprietary models like GPT-4o show strong integration, while most open-source systems lag, particularly in culturally diverse and scientific scenarios. Our analysis further identifies common failure modes, including semantic hallucination and the underrepresentation of non-Western knowledge, underscoring the limitations of current models in achieving real-world readiness. These findings highlight the need to guide future T2I models toward deeper, knowledge-aware generation. Project Page: \href{https://dwanzhang-ai.github.io/WorldGenBench/}{https://dwanzhang-ai.github.io/WorldGenBench/}

cs.CV↗

VideoWeaver: Evaluating and Evolving Skills for Agentic Long Video Generation

Agentic long video generation requires planning, tool orchestration, and cross-clip coordination over a long horizon. Most existing video agents either rely on static, human-crafted workflows, which require substantial manual effort and poorly adapt across tasks, or iteratively refine the output of the current task without persistently distilling execution experience into reusable skills for future tasks. We introduce VideoWeaver, an agent harness and benchmark that evaluates and evolves skills for long video generation. Given a single high-level instruction, an agent dynamically composes foundation skills into its own workflow rather than following a predefined pipeline. We construct a benchmark of 16 task categories and 285 cases, with references spanning text, image, audio, video, and their combinations. We further propose an evidence-grounded agent-as-judge that inspects both the execution trace and the final video to diagnose process and output failures. Based on this feedback, our evolution algorithm progressively refines category-level composition and creator skills, allowing recurring experience to guide dynamically constructed workflows for unseen cases. Experiments show that explicit composition skills improve the generation process over foundation skills alone, while skill evolution further improves output quality and generalizes to unseen cases. Incorporating judge feedback yields additional gains, especially on output metrics, and the agent-as-judge aligns well with human, particularly on process metrics. Code is available at https://github.com/JianhuiWei7/VideoWeaver.

cs.CV↗

Dependency-Aware Reward Shaping for Agentic Reinforcement Learning

When training large language models with reinforcement learning, terminal rewards provide little guidance about which steps matter. Common methods for assigning step credit overlook that work built on uncorrected mistakes is wasted while independent work remains valid. With only a final success/failure reward, every step in a failed episode has zero total future reward, even when it made progress. We propose Dependency-Aware Reward Shaping (DARS), which represents task progress as predicates linked by prerequisite relations and assigns step-level credit over the dependency graph. An annotator marks which predicates each step verifies, invalidates, or repairs. Verified predicates are discounted according to graph distance from the nearest broken prerequisite, while independent predicates are unaffected. Repairs update these weights based on any errors that remain; invalidated predicates need re-verification to regain credit. A fixed potential converts these annotations into signed per-step rewards. A common reward and annotation interface allows DARS to integrate with a range of reasoning and agentic training methods, such as GiGPO and ARPO/AEPO, without changing their rollout strategies or optimizers. Across five task families and models from 1.5B to 8B, DARS improves success by up to 10 points over GiGPO trained with the same budget and harness (ALFWorld), raises the WebShop task score and Search-R1 QA accuracy, complements AEPO's entropy-based training on AIME24/25 with a Python interpreter, and exceeds OmniOPD in controlled tool-free reasoning comparisons at 1.7B and 4B. Ablations show that step-level credit, dependency attenuation, and graph topology each contribute. On ALFWorld, a distilled 8B annotator matches the API annotator, enabling DARS to run efficiently without a frontier judge. Code is available at https://github.com/JianhuiWei7/DARS.

cs.AI↗

LIBERO-MAX: Do Robot Policies Adapt When the World Changes?

Robots must often continue a task after a target moves, the viewpoint shifts, or an obstacle appears, even though their earlier observations and committed actions reflect the previous scene. Many simulation robustness benchmarks fix external conditions at reset, leaving this temporal challenge underexamined. We introduce LIBERO-MAX, a benchmark of 8,000 paired cases spanning eight types of changes to geometry, observations, appearance, clutter, and paths. Each pair compares task execution with and without a mid-task event, holding the task, initial state, policy seed, and pre-event action sequence fixed. This controlled comparison distinguishes event-associated regressions from failures already present without the change. Across fourteen current VLA, hybrid, and world-action policies, events reduce success by 11.0-25.7 percentage points. Event profiles reveal shared vulnerabilities to geometry and observation changes, while policy-family rankings interleave. Camera controls show that robustness reflects both competence under the changed conditions and the trajectory from which they are encountered; varying query cadence does not eliminate the gap. Together, the paired protocol and temporal diagnostics establish LIBERO-MAX as a reproducible testbed for diagnosing failures under mid-execution changes and measuring progress toward robot policies that remain effective as the world changes.

cs.RO↗

Complementary Retrieval-Augmented Prompting for Consistent Long-Form Video Generation

While recent video foundation models excel at generating high-quality short videos, long-form video generation remains a critical challenge, where a major bottleneck lies in conditioning independently generated shots to preserve consistent characters, scenes, and objects throughout a story. Existing training-free approaches typically condition target shots using retrieved historical visuals. However, these references often suffer from severe informational mismatch, either introducing irrelevant contextual redundancy or failing to provide the full combination of required elements for the target shot. To resolve this, we present Complementary Retrieval-Augmented Prompting, an agentic framework that strategically aggregates a compact set of mutually supportive historical references to achieve complete and targeted conditioning for long-form video generation without retraining or modifying the underlying generator. Specifically, our framework explicitly models the visual elements required by each target shot by parsing the narrative script into a text-grounded visual element registry that tracks characters, objects, scenes, and their shot-level states. A VLM-annotated keyframe library further maps these elements to past visual observations. Guided by the required elements, our agent retrieves complementary references that maximize target-element coverage while minimizing historical noise. Finally, the retrieved references, structured element states, and grounding instructions are assembled into a unified prompt for the frozen video generator. This element-aware process provides comprehensive conditioning while remaining fully interpretable. Quantitative and qualitative evaluations on multi-shot story generation demonstrate that our method consistently outperforms recent-frame, memory-based, and entity-level retrieval baselines in cross-shot consistency and text-controllability.

cs.CV↗

Beyond Oracle Communication: Benchmarking Interactive Intent Alignment Under Miscommunication and Evolving User Intent

Modern LLM agents increasingly tackle complex tasks through interactive, long-horizon exchanges with users, while existing benchmarks generally assume that users always accurately and sufficiently communicate a fixed intent. However, this oracle communication assumption rarely holds in practice: users may miscommunicate, change their goals, and run out of patience. We define this task setting as Interactive Intent Alignment, where agents must recover and continuously track the user's current intent despite imperfect communication and evolving goals. To study this setting, we introduce Drift-Bench++, a principled benchmark construction pipeline for verified executable tasks with controlled misalignment and intent shifts, along with an interaction protocol featuring finite patience, diverse simulated users, and silent interaction-conditioned shifts. We further develop GRIP, a comprehensive evaluation protocol covering task grounding, user realism, inquiry effectiveness, and adaptation to evolving intent. Across diverse environments, models, and interaction conditions, stronger interaction consistently helps but remains far from oracle performance; Validation on deployed ProdAgent sessions further shows that the modeled failures are prevalent and consequential in deployment. By providing a unified, executable benchmark for interactive intent alignment, Drift-Bench++ offers a foundation for evaluating and advancing agents under realistic communication and evolving intent.

cs.CL↗

TraceDance: An Automated System for Building Agent Behavior Benchmarks from Real-World Agent Deployment Traces

An agent can complete a task while exhibiting undesirable behavior during execution. Developers need tests for the specific behaviors encountered in deployment, beyond fixed benchmark suites. We present TraceDance, an agent system that constructs targeted benchmarks from deployment traces for user-specified undesirable behaviors. For efficient construction, Anchor-and-Confirm combines programmable retrieval with candidate-level confirmation by a Flash large language model (LLM), while the Anchor Synthesis Loop generates and revises specifications for custom behaviors. The benchmarks use decision-point continuation to evaluate an LLM's next turn at a recorded decision point with a behavior-specific rubric, without a reference answer or environment replay. Experiments in coding and general tool use draw on 252,557 sessions and produce 107 benchmarks with 4,125 instances, fulfilling 95.3% of build-target requests. Both human annotators confirm the requested behavior in 84% of sampled instances, and the automated grader's agreement with human pass/fail judgments is comparable to that between the annotators. Nine frontier LLMs achieve a mean pass rate of only 26.7%, showing that they still struggle to respond appropriately at the evaluated decision points. Analysis across behavior-specific benchmarks further reveals weaknesses in how current LLMs behave as agents. By turning deployment problems into targeted benchmarks, TraceDance could serve as a key component of the recursive self-improvement (RSI) loop.

cs.AI↗

Self-Boosting Vision-Language Models with Noisy Student On-Policy Self-Distillation

Post-training enables vision-language models (VLMs) to understand human instructions and perform various downstream tasks. Current post-training methods usually rely on human-annotated data, distillation from external models, reinforcement learning with human feedback, or verifiable answers. This limits their ability to improve without external supervision. To tackle this, we propose NOPD (Noisy Student On-Policy Self-Distillation), a simple yet effective self-distillation approach that improves VLMs without any external models or ground-truth answers. Our key insight is that prediction discrepancies between clean and corrupted inputs naturally induce a self-supervision signal. In NOPD, the model learns from corrupted inputs while using its own predictions under clean inputs as token-level supervision. We show the effectiveness of NOPD on five visual reasoning tasks; it can match and even outperform reinforcement learning approaches or distillation from external models. Notably, when trained with 2.1K samples from Geometry3K, NOPD improves Qwen2.5-VL-7B by 20 points on its validation set. It also shows generalization on out-of-distribution test sets and achieves 7.4 point gains on MathVista. Furthermore, we demonstrate that NOPD is a general approach to enhance VLMs, achieving improvements across three models on 12 benchmarks.

cs.LG↗

NTIRE 2024 Challenge on Image Super-Resolution (x4): Methods and Results

This paper reviews the NTIRE 2024 challenge on image super-resolution ($\times$4), highlighting the solutions proposed and the outcomes obtained. The challenge involves generating corresponding high-resolution (HR) images, magnified by a factor of four, from low-resolution (LR) inputs using prior information. The LR images originate from bicubic downsampling degradation. The aim of the challenge is to obtain designs/solutions with the most advanced SR performance, with no constraints on computational resources (e.g., model size and FLOPs) or training data. The track of this challenge assesses performance with the PSNR metric on the DIV2K testing dataset. The competition attracted 199 registrants, with 20 teams submitting valid entries. This collective endeavour not only pushes the boundaries of performance in single-image SR but also offers a comprehensive overview of current trends in this field.

cs.CV↗

How Far Are Video Models from True Multimodal Reasoning?

Despite remarkable progress toward general-purpose video models, a critical question remains unanswered: how far are these models from achieving true multimodal reasoning? Existing benchmarks fail to address this question rigorously, as they remain constrained by straightforward task designs and fragmented evaluation metrics that neglect complex multimodal reasoning. To bridge this gap, we introduce CLVG-Bench, an evaluation framework designed to probe video models' zero-shot reasoning capabilities via Context Learning in Video Generation. CLVG-Bench comprises more than 1,000 high-quality, manually annotated metadata across 6 categories and 47 subcategories, covering complex scenarios including physical simulation, logical reasoning, and interactive contexts. To enable rigorous and scalable assessment, we further propose an Adaptive Video Evaluator (AVE) that aligns with human expert perception using minimal annotations, delivering interpretable textual feedback across diverse video context tasks. Extensive experiments reveal a striking answer to our central question: while state-of-the-art (SOTA) video models, such as Seedance 2.0, demonstrate competence on certain understanding and reasoning subtasks, they fall substantially short with logically grounded and interactive generation tasks (achieving success rates <25% and ~0%, respectively), exposing multimodal reasoning and physical grounding as critical bottlenecks. By systematically quantifying these limitations, the proposed method provides actionable feedbacks and a clear roadmap toward truly robust, general-purpose video models. CLVG-Bench and code are released here.

cs.CV↗

A Versatile Multimodal Agent for Multimedia Content Generation

With the advancement of AIGC (AI-generated content) technologies, an increasing number of generative models are revolutionizing fields such as video editing, music generation, and even film production. However, due to the limitations of current AIGC models, most models can only serve as individual components within specific application scenarios and are not capable of completing tasks end-to-end in real-world applications. In real-world applications, editing experts often work with a wide variety of images and video inputs, producing multimodal outputs -- a video typically includes audio, text, and other elements. This level of integration across multiple modalities is something current models are unable to achieve effectively. However, the rise of agent-based systems has made it possible to use AI tools to tackle complex content generation tasks. To deal with the complex scenarios, in this paper, we propose a MultiMedia-Agent designed to automate complex content creation. Our agent system includes a data generation pipeline, a tool library for content creation, and a set of metrics for evaluating preference alignment. Notably, we introduce the skill acquisition theory to model the training data curation and agent training. We designed a two-stage correlation strategy for plan optimization, including self-correlation and model preference correlation. Additionally, we utilized the generated plans to train the MultiMedia-Agent via a three stage approach including base/success plan finetune and preference optimization. The comparison results demonstrate that the our approaches are effective and the MultiMedia-Agent can generate better multimedia content compared to novel models.

cs.CV↗