arXiv subjects

Dakuo Wang

Publications and source records attributed to Dakuo Wang.

At least 19 recordsLinked to original sources

arXiv2026

Harness Updating Is Not Harness Benefit: Disentangling Evolution Capabilities in Self-Evolving LLM Agents

LLM agents are increasingly deployed as systems built around editable external harnesses, including prompts, skills, memories and tools, that shape task execution without changing model parameters. Harness self-evolution adapts such agents by updating these harnesses from execution evidence. Yet it remains unclear whether a model's base capability in task-solving predicts its capabilities in harness self-evolution: which models produce useful harness updates, and which actually benefit from them? We analyze two harness self-evolution capabilities: (i) harness-updating, the capability to produce useful persistent harness updates from execution evidence; (ii) harness-benefit, the capability to benefit from updated harnesses during task solving. Our analysis reveals two findings. First, harness-updating is flat in base capability: models from different capability tiers produce harness updates that lead to surprisingly similar gains; even Qwen3.5-9B's updates yield gains comparable to those of Claude Opus~4.6. Second, harness-benefit is non-monotonic in base capability: weak-tier models benefit little from updated harnesses, mid-tier models benefit most, and strong-tier models benefit less than mid-tier. We trace low gains at the weak tier to two failure modes: weak-tier models may fail to activate relevant harness artifacts, or activate them but fail to follow them faithfully. These findings suggest investing capability budget in the task-solving agent rather than the evolver, and targeting harness invocation and long-horizon instruction following in agent training. Our source code is publicly available at https://github.com/A-EVO-Lab/a-evolve/tree/release/harness-evolution.

cs.AI↗

arXiv2026

SpanUQ: Span-Level Uncertainty Quantification for Large Language Model Generation

Uncertainty estimation is essential for the trustworthy deployment of LLMs: it underpins self-refinement and serves as a decision signal for downstream systems, from verifiers to collaborating agents. However, existing approaches operate at suboptimal granularities: token-level scores lack semantic coherence, while sequence-level scores fail to localize errors. We formalize Span-Level Uncertainty Estimation (SLUE), a new task that targets the natural granularity for uncertainty: semantically coherent text spans, each conveying a single assessable unit of meaning. To address this task, we introduce SPANUQ, a lightweight probe that distills the uncertainty knowledge from expensive multi-sample inference into a single forward pass over LLM hidden states. SPANUQ employs a DETR-style span decoder to simultaneously detect spans and estimate their uncertainty via a Mixture of Beta distribution, trained with a principled combination of Beta NLL regression and contrastive ranking objectives. We construct SPANUQ-BENCH, the first span-level uncertainty benchmark comprising 20K prompts, 293K annotated spans, and continuous soft labels derived from multi-sample claim verification. Experiments on five LLM backbones show that SPANUQ consistently achieves the best span-level uncertainty quality , outperforming the strongest probe baseline and all sampling-based methods while being 10~20x faster. Its DETR-based span detector attains 0.910 F1, surpassing the best heuristic by 39.4%, enabling precise error localization that sequence-level methods cannot provide. The same architecture ports to five LLMs spanning two model families, with one probe trained per backbone, and we additionally observe that sequence-level uncertainty is partially decomposable, suggesting that span-level estimation subsumes sequence-level as a special case. The project page is available damon-demon.github.io/SpanUQ.

cs.CL↗

arXiv2026

Language Models as AI Research World Models

AI research agents automate the cycle of proposing, implementing, and evaluating experiments, opening a path toward recursive self-improvement. Yet their ability to propose experiments outpaces their capacity to execute them in real environments, making outcome prediction a key capability for sustained self-improvement under limited experimental budgets. We investigate language models as Research World Models (RWMs), which predict the outcomes of candidate interventions across research environments. Our evaluation draws on over 2,600 experimental records from nine research environments spanning pretraining, post-training, and inference, representing more than 171,000 H100 GPU-hours of experimentation. Research knowledge acquired from real experimental experience improves RWM predictions of unseen interventions within the same environment (Spearman +0.27), and can be reused across environments. For example, using only pretraining experience from OLMo3, Marin, and Nanochat, an RWM reduces selection regret in the Qwen3 environment by 78% compared with zero-experience setting. These benefits extend to multi-round Autoresearch under a fixed selection budget: RWMs with in-env and cross-env research knowledge increase the best gain achieved by 15.8% and 11.6%, respectively. Ablations across 13 language models used as RWMs show that adding research knowledge can improve intervention ranking more than changing models or increasing reasoning effort alone. These findings support language models as RWMs and motivate accumulating experimental data for future RWM training.

cs.CL↗

arXiv2026

How Do AI Agents Waste Your Money?

AI agents may consume massive tokens across a wide range of tasks, but how many of those tokens are really necessary? Although prior work has shown that agents can incur unexpectedly high token costs, we still lack a systematic way to identify and measure token waste. In this work, we introduce a unified framework for measuring token waste across heterogeneous agent systems. Our framework records and normalizes complete agent trajectories, identifies removable spans using an LLM-as-a-Judge (LLMaJ), and maps the identified text spans back to the original trajectory for token-level measurement. We validate the identified waste through direct interventions by removing marked spans, resuming agent execution, and measuring changes in task outcomes and subsequent actions. In online pruning experiments, we remove 41-46% of the model's reasoning while largely preserving task performance across two distinct agent benchmarks. Using this framework, we analyze agents across multiple settings. Our analysis shows that token waste has a system-level structure that aggregate token usage alone cannot reveal. First, waste varies more across tasks and agent settings than across models. Second, reasoning and visible execution can exhibit different waste patterns. Third, different harnesses can introduce distinct sources of inefficiency even under the same model and task. Overall, our framework transforms token waste into an actionable optimization signal, enabling researchers and practitioners to train more efficient models, design better reasoning strategies, and optimize agent harnesses toward reducing unnecessary computation rather than merely reducing token usage.

cs.CL↗

arXiv2026

Adaptive Auto-Harness: Sustained Self-Improvement for Agentic System Deployment on Open-Ended Task Streams

Auto-harness systems improve LLM agents by adapting prompts, skills, tools, and supporting infrastructure from execution feedback. Sustaining these gains on open-ended task streams requires handling growing histories, heterogeneous tasks, and shifting distributions. We introduce Adaptive Auto-Harness, a unified framework that separates persistent evolution workspaces from deployable harnesses and supports interchangeable evolution procedures. We compare configurations adapted from five prior approaches on recorded streams of prediction markets, security challenges, and event forecasts. With the same Sonnet solver, our full system reaches 80.9\% market all-task accuracy and 50.2\% security Pass@1, versus 22.2\% and 37.2\% without evolution; market accuracy also exceeds a market-following control's 72.0\%. Our analyses show that artifact accumulation does not guarantee sustained gains, and knowledge that transfers across tasks can still lose value as the task distribution changes. Continued evolution can therefore help preserve earlier gains through revision of accumulated knowledge. Code: https://github.com/A-EVO-Lab/a-evolve.

cs.LG↗

arXiv2026

Humans' ALMANAC: A Human Collaboration Dataset of Action-Level Mental Model Annotations for Agent Collaboration

Recent advances in LLM agents have enabled complex cognitive capabilities, such as multi-step reasoning, planning, and tool use, that increasingly position these agents as collaborative partners in human workflows. Effective collaboration, however, requires collaborators to continuously maintain and align mental models of their own reasoning, partners' intentions, and shared goals during the collaborative process. Today's agents rarely develop such capabilities since they are primarily optimized for task completion rather than for maintaining the underlying mental models required for collaborative processes. Moreover, the community lacks authentic human collaboration data with action-level mental model annotations that could guide agents toward collaborative competence. To bridge this gap, we present ALMANAC, a dataset of Action-Level Mental model ANnotations for Agent Collaboration built from the Map Task, a classic dyadic routing task from social science. ALMANAC contains 2,987 collaboration actions, each paired with theory-informed mental model annotations that record the participants' self-reasoning, perceived partner intent, and perceived team goal. We benchmark six LLMs on predicting humans' next-turn behavior and mental models. Our results demonstrate ALMANAC's utility in evaluating models' ability to simulate human collaborative behaviors and infer their underlying mental models. The project page is available at https://human-agent-collab.hailab.io/almanac.

cs.AI↗

arXiv2026

Firefly: Illuminating Large-Scale Verified Tool-Call Data Generation from Real APIs

Training tool-calling agents requires large-scale trajectory data with verifiable labels, yet existing approaches either synthesize environments that diverge from real API behavior or generate tasks without ground-truth outcomes for verification. We present FireFly, a pipeline for generating verified tool-call data from real-world MCP servers. Our key insight is to invert the standard synthesis pipeline: rather than generating tasks and hoping they are solvable, we first let a strong LLM explore real APIs along graph-guided DAG structures, then synthesize tasks backward from observed outcomes, guaranteeing label correctness by construction. To handle the scale of real-world tool spaces (${\sim}$1,000 tools), we build a pairwise tool graph and sample sub-DAGs to focus exploration on semantically coherent workflows. To address environment drift in live APIs, we construct a retrieval-augmented simulator that caches all exploration results and replays them during training and evaluation, enabling fully offline and reproducible RL. Applying this pipeline yields 5,144 verified tasks spanning 240 servers and 993 tools. A 4B-parameter model trained with GRPO on FireFly matches Claude Sonnet 4.6 on our held-out test set and shows improvements on multiple tool-calling benchmarks including Tau2-Bench, MCPMark, and MCP-Atlas.

cs.SE↗

arXiv2026

StaRPO: Stability-Augmented Reinforcement Policy Optimization

Reinforcement learning (RL) is effective in enhancing the accuracy of large language models in complex reasoning tasks. Existing RL policy optimization frameworks rely on final-answer correctness as feedback signals and rarely capture the internal logical structure of the reasoning process. Consequently, the models would generate fluent and semantically relevant responses but logically inconsistent, structurally erratic, or redundant. To this end, we propose StaRPO, a stability-augmented reinforcement learning framework that explicitly incorporates reasoning stability into the optimization objective. Our StaRPO decomposes stability into two computable lightweight metrics: the Autocorrelation Function (ACF) to evaluate local step-to-step coherence, and Path Efficiency (PE) to evaluate global goal-directedness of the reasoning trajectory. These stability rewards are combined with task rewards to provide complementary and process-aware feedback. We validate the effectiveness of using ACF and PE rewards by showing their correlation with logic errors on two backbone models. Experiments on four reasoning benchmarks show that StaRPO consistently outperforms compared baselines and can enhance both final-answer accuracy and logical stability.

cs.AI↗

arXiv2026

MiniCorp: The Last Mile to an AI Agent Corporation

The last mile toward enterprise AGI is a company that runs itself. Training and adapting such agents require longitudinal enterprise data, which remain scarce, costly to acquire, and often restricted by privacy constraints. Historical archives are also frequently incomplete and record only what actually happened. They cannot show the outcomes of alternative decisions. We introduce MiniCorp, an office simulator for studying how agents can collectively run a company while generating enterprise data at scale. Using an e-commerce company as a demonstration, MiniCorp connects two interacting worlds. The external world models customers, dynamic competitors, and market mechanisms. The internal world consists of agents that observe events, discuss their options, and make strategic decisions. These decisions have lasting effects on the market, and the resulting feedback informs the firm's later decisions. As the firm and market interact, MiniCorp continuously records the agents' communications and decisions. These records preserve the information available at the time and the business results that followed. Checkpointing allows the same situation to be replayed under different decisions, providing comparisons unavailable in static archives. We evaluate end-to-end fidelity against patterns reported in empirical studies of real markets. These evaluations provide agents with realistic market feedback and reduce the risk that they learn to exploit flaws in the simulator. Our experiments show agents coordinating across roles and adapting their decisions to market feedback. With explicit long-term strategic guidance, they also sustain advertising exploration despite weak early returns. MiniCorp thus provides an environment for studying AI-run companies and a scalable source of longitudinal and counterfactual enterprise data for agent training and evaluation.

cs.AI↗

arXiv2026

Bench2Robust: Scenario-Controlled Recovery for Tool-Using Agents

Tool-using LLM agents are typically developed and post-trained in controlled environments with reliable tools, but deployment exposes them to tool failures, API and schema drift, and third-party services outside the developer's control. We introduce Bench2Robust, a framework for scenario-controlled solvability that augments existing benchmarks with episodes requiring retrying the original path, switching to a fallback-equivalent tool, or recognizing that no viable path remains. By controlling recoverability, Bench2Robust separates recovery-policy quality from randomness in whether failures persist or resolve. Across 7 models from 4 families and 10 task/domain slices spanning customer-service and function-calling benchmarks, 69 of 70 model-subset pairs degrade under tool failures, by up to 46.7 percentage points. We propose Bayesian Tool Memory (BTM) for structured runtime recovery context and use reinforcement learning (RL) for learned recovery behavior. On held-out Retail tasks using the benchmark's original tool set, BTM improves robustness by 16.8 percentage points without retraining, while RL retains a 6.3-point gain without inference-time BTM. Combining RL with BTM yields a 20.7-point improvement over the base model while preserving clean performance. These results show that robust tool use benefits from combining structured runtime recovery context with learned recovery behavior.

cs.AI↗

arXiv2026

Deco: Extending Cherished Physical Objects into AI Companion Agents through Dual Embodiment

Physical objects (e.g., plush toys) can transcend materiality to become emotional anchors and provide companionship. However, these bonds remain one-sided because most physical objects cannot reciprocate. AI companions offer responsiveness and personalization, but typically entail building bonds from scratch. We investigated how AI companions might inherit and extend users' existing bonds with physical objects. A formative study (N=9) informed four design principles (Faithful Identity, Calibrated Agency, Ambient Presence, Reciprocal Memory), shaping our Dual-Embodiment Companion Framework. We instantiated it as Deco to create digital embodiments of physical companions. In a within-subjects lab study (N=25), Deco was rated higher than a personalized digital-only companion on six companion-related measures (all p<.01). A subsequent seven-day field deployment (N=17) showed sustained engagement, higher post-deployment well-being (p=.040), and three key relational patterns: digital activities retroactively vitalized physical objects, bond deepening centered on emotional engagement depth, and participants sustained bonds while navigating companions' AI nature. Dual embodiment offers a promising framework for revitalizing physical objects with AI agents.

cs.HC↗

arXiv2026

Evo-Harness: Context-to-Harness Skill Compilation for Self-Evolving Agents

Learning from experience is critical for developing capable, self-improving large language model (LLM) agents. Existing methods typically extract knowledge from accumulated trajectories via reflection, memory, rules, or skills. However, agents in realistic environments continuously encounter novel tasks, often offering only a one-shot opportunity to improve. These executions yield rich but highly noisy contexts, entangling broadly useful lessons with task-specific artifacts. Critically, prior works rarely validate their effectiveness on complex real-world tasks or isolate the underlying drivers of improvement. To address these gaps, we formulate online harness learning, where a frozen agent improves by continually updating a structured harness across sequential tasks. This formulation enables a systematic study of key self-improvement factors through our proposed Evo-Harness. At its core, context-to-harness skill compilation distills noisy, single-shot executions into reusable skill harnesses for cross-domain and topic-level adaptation. To demonstrate the efficacy of one-shot skill compilation, we evaluate across five realistic benchmarks (TerminalBench2, SWE-bench, CL-Bench, -bench, WebArena-Infinity). Our extensive analysis demonstrates the effectiveness of Evo-Harness and provides a principled understanding of how LLM agents can effectively learn on the fly. Our code is available at https://github.com/A-EVO-Lab/a-evolve/tree/release/evo-harness.

cs.AI↗

arXiv2026

PACEShop: Evaluating Personalized, Actionable, Compositional, and Evidence-grounded Shopping Assistants

Shopping assistants are shifting from ranked product lists toward structured decision support, where systems must synthesize shopper context, product evidence, and next-step guidance into a coherent recommendation experience. This changes the unit of evaluation: a fluent response can still fail by ignoring shopper context, contradicting itself across components, or leaving defects too vague to localize. Existing personalization, grounding, and LLM-as-a-judge benchmarks cover pieces of this problem, but they do not define a joint evaluation target for structured shopping-assistant responses. We formulate this missing evaluation target as PACE: Personalized, Actionable, Compositional, and Evidence-grounded evaluation. We instantiate PACE with two artifacts: PACEShop, a benchmark dataset that makes the target measurable through 22,625 controlled records with structured personas, auditable evidence pools, GOOD/BAD labels, and gold defect family and location annotations; and PACEJudge, a training-free judging protocol that makes the target reportable through a structured output contract. Our experiments show that generic judges can recognize broad quality but fail to recover the diagnostic fields required for PACE; PACEShop makes these failures verifiable, and PACEJudge improves persona-source, cross-component, grounding, and family/location closure without retraining, showing that realistic shopping-assistant evaluation requires a task-matched output contract rather than only a stronger backbone or scalar prompt.

cs.CL↗

arXiv2026

PADFormer: Pose-agnostic Anomaly Detection from Sparse View Images

Pose-agnostic Anomaly Detection (PAD) remains challenging as anomalies can appear under arbitrary viewpoints, requiring methods to handle significant pose variations. Existing approaches rely on complex 3D reconstruction, which are computationally expensive and require extensive multi-view data. We propose PADFormer, a novel image-space approach that leverages Vision Transformer (ViT) to directly reconstruct anomaly-free versions of query images while preserving pose information. Our key insight is to adapt cross-view masked reconstruction for anomaly detection through training exclusively on normal data, combined with dynamic patch selection and spatial alignment mechanisms that enable effective learning from sparse reference views under significant pose variations. During inference, we perform multiple forward passes with different masking patterns to generate an ensemble of anomaly-free reconstructions, ensuring comprehensive coverage of the query image. Anomalies are detected by comparing these reconstructions with the query image. PADFormer achieves state-of-the-art results on the PAD benchmark while maintaining comparable performance on classic few-shot anomaly detection (FSAD) tasks, demonstrating superior efficiency and generalization without requiring 3D reconstruction.

cs.CV↗

arXiv2026

Seeing Through the Forecast Clutter: Communicating Climate Forecast Distributions with Weighted Multiple Forecast Visualizations

Forecasts often diverge because different models make varying assumptions to account for underlying uncertainty. Readers who consume forecasts may wish to survey the shape and spread of these multiple forecasts to get a full account of the different predictions. One approach to visualizing multiple forecasts is through Confidence Interval (CI) plots. However, while the summative CI plots can communicate uncertainty of an ensemble, they obscure attributes of individual forecasts that can lead to inaccurate perceptions of the distribution of these forecasts (e.g., implying a normal distribution when non-existent). To address this challenge, we investigate the use of multiple forecast visualization (MFV) in communicating nuanced forecast distributions through two preregistered experiments using climate forecast data. In Experiment 1 (480 participants), we compared how well MFV and CI plots can represent the distribution of multiple forecasts. We found that, compared to CI plots, MFV improved participants' ability to identify the underlying distribution of forecasts and reduced the likelihood of assuming normality. Building on Experiment 1, we examined in Experiment 2 (900 participants) whether a downsampled MFV showing 9 forecasts might be able to communicate additional forecast properties using linewidth and opacity without negatively impacting distribution perception. We found that visually weighting forecasts by linewidth or opacity preserves readers' perception of the underlying distribution. We discuss how these findings suggest the use of downsampled and weighted MFV to cut through forecast clutter by aligning perceived distribution with the underlying forecast distribution, while opening up design opportunities to use weighting to communicate additional forecast attributes.

cs.HC↗

arXiv2026

Toward Metaphor-Fluid Conversation Design for Voice User Interfaces

Metaphors play a critical role in shaping user experiences with Voice User Interfaces (VUIs), yet existing designs often rely on static, human-centric metaphors that fail to adapt to diverse contexts and user needs. This paper introduces Metaphor-Fluid Design, a novel approach that dynamically adjusts metaphorical representations based on conversational use-contexts. We compare this approach to a Default VUI, which characterizes the present implementation of commercial VUIs commonly designed around the persona of an assistant, offering a uniform interaction style across contexts. In Study 1 (N=130), metaphors were mapped to four key use-contexts-commands, information seeking, sociality, and error recovery-along the dimensions of formality and hierarchy, revealing distinct preferences for task-specific metaphorical designs. Study 2 (N=91) evaluates a Metaphor-Fluid VUI against a Default VUI, showing that the Metaphor-Fluid VUI enhances perceived intention to adopt, enjoyment, and likability by aligning better with user expectations for different contexts. However, individual differences in metaphor preferences highlight the need for personalization. These findings challenge the one-size-fits-all paradigm of VUI design and demonstrate the potential of Metaphor-Fluid Design to create more adaptive and engaging human-AI interactions.

cs.HC↗

arXiv2026

SENTINEL: Failure-Driven Reinforcement Learning for Training Tool-Using Language Model Agents

Language model agents are increasingly effective in solving realistic tasks through multi-turn tool use. However, training reliable tool-using agents remains challenging in practice. While reinforcement learning provides an on-policy paradigm for improving agents from their own environment interactions, its effectiveness depends heavily on the training task distribution. When tasks are fixed before training, the task distribution can become increasingly mismatched with the policy's evolving capabilities, causing many rollouts to be spent on uninformative tasks. We propose SENTINEL, a failure-driven reinforcement learning framework that turns the Solver's rollout failures into targeted training tasks. SENTINEL follows a Controller--Proposer--Solver loop: the Controller analyzes failed trajectories and summarizes recurring error patterns, the Proposer generates executable tasks that stress these weaknesses, and the Solver is trained on the targeted tasks. On Tau2-Bench Retail with Qwen3-4B-Thinking-2507, SENTINEL improves Pass\^{}1 from 66.4 to 74.9 and outperforms RL on general synthetic tasks across Pass\^{}k metrics. These results demonstrate that model failures provide an effective and scalable source of targeted training signal for improving tool-using language model agents.

cs.CL↗

arXiv2026

CollabSim: A CSCW-Grounded Methodology for Investigating Collaborative Competence of LLM Agents through Controlled Multi-Agent Experiments

Multi-agent systems (MAS) built on large language models have shown growing promise, with their effectiveness resting on agents' ability to coordinate through text-based channels much as human teams do. Yet recent study suggests that MAS often falter not because agents lack individual task-solving ability, but because they lack collaborative competence: the capacity to establish common ground, maintain shared task understanding, balance individual and collective incentives, and repair misalignment as interaction unfolds. Decades of research in Computer-Supported Cooperative Work have characterized these requirements for human teams coordinating under constrained communication, yet existing MAS evaluations focus mainly on task outcomes or single-agent proficiency in reasoning, planning, and tool use. To enable a systematic analysis of agents' collaborative competence in MAS, we introduce CollabSim, a configurable simulation framework that combines a theory-grounded definition of collaborative capabilities, controlled manipulation of interaction conditions, and action-level probing of agents' internal states. Experiments across four LLMs show that CollabSim can capture condition effects, separate model performance patterns, and reveal task-dependent effects of agent design.

cs.CL↗