Search arXiv⌕ Search

arXiv subjects

Chien-Sheng Wu

Publications and source records attributed to Chien-Sheng Wu.

At least 19 recordsLinked to original sources
arXiv2026

MTA-Zero: Self-Improving Multimodal Deep Search via Online QA-Solver Co-training

Multimodal large language models (MLLMs) have become strong visual reasoners, yet remain limited in open-world, multi-step search that requires grounding images, selecting tools, and integrating web evidence. Existing approaches typically construct and filter fixed offline corpora whose question distributions cannot adapt as the target deep-search agent improves. Building on MTA-Agent's image-grounded QA construction, we present MTA-Zero, an online multimodal QA-Solver co-training framework for self-improving multimodal deep-search agents that requires no precomputed multi-hop corpus. Starting from image-grounded VQA seed contexts, a trainable QA Generator retrieves evidence and constructs verified multi-hop questions online. A separate deep-search Solver then selects tools, gathers evidence, and answers the generated questions. To focus generation on the Solver's learning frontier, we derive a Goldilocks reward for the QA Generator from the Solver's group-wise rollout success rate: questions solved by all rollouts or by none receive zero reward, whereas questions near the current frontier are favored. We optimize the two policies independently with GRPO, without sharing parameters or gradients. Across six multimodal deep-search benchmarks, MTA-Zero improves a Qwen3-VL-8B Solver from 36.91% to 51.58% mean accuracy and achieves performance comparable to GPT-5 under the same agent workflow. The same framework also improves a newer Qwen3.6-27B Solver by 10.82 points. These results show that verified online question generation provides a policy-specific curriculum for continually improving multimodal deep-search agents.

cs.CL↗
arXiv2026

Training Reasoning Models to Generate Diverse Scientific Ideas

One of the first steps of a research pipeline is generating ideas. Large Language Models (LLMs) are increasingly used to automate this step, but most of this work optimizes the quality of individual ideas, while overlooking diversity. In our work, we study document-grounded ideation, focusing on increasing the diversity of LLM generated ideas and ensuring they are high quality. Given prior papers, a model must propose a set of ideas that are each high-quality and clearly distinct from one another, while also being grounded in the input documents. To optimize for this, we propose training with a set-level reinforcement learning objective. In our setup, in a single generation, the model writes K structured idea cards grounded in the source papers, and the set is scored jointly, where each idea's quality is weighted by its distance from the ideas written before it. This teaches the model to steer each new idea away from its predecessors. We show that our trained 8B model achieves higher creative utility (a measure combining quality and diversity) than Claude Opus 4.8 and GPT-6 Astra on our contamination-free benchmark. In a pilot human study with four experts, our sets are judged as promising as Claude Opus 4.8's and contain marginally more unique ideas. When a coding agent runs proof-of-concept experiments on the ideas, our model's success@k is comparable to Claude Opus 4.8's, so we see that the added diversity does not cost any executability.

cs.CL↗
arXiv2026

Self-Retrospection Distillation: Turning Post-hoc Experiences into Prior Foresight

Reinforcement learning with verifiable rewards (RLVR) turns agent experience into learning signals primarily through scalar outcome rewards after interaction. For group-relative objectives, however, this signal vanishes when all rollouts receive the same reward, even though their trajectories may reveal useful information about what the task requires and how the agent fails. We ask a complementary question: can hindsight teach an agent what it could have anticipated before acting? We introduce prospective learning, which uses post-hoc experience to supervise foresight predictions from the pre-interaction view, and instantiate it with Self-Retrospection Distillation (SRD). Intuitively, a completed trajectory reveals knowledge that would have been useful and pitfalls that should be avoided; SRD distills this privileged hindsight into trajectory-blind foresight of the same policy. Foresight serves only as a training target and need not be explicitly generated at inference time. Across 10 tool-integrated reasoning and long-horizon agentic tasks, SRD complements RLVR and self-distillation baselines with gains of up to 24.2 pp. Its advantage is especially pronounced when reward contrast is scarce: when 37--98% of rollout groups are reward-uniform across model scales, yet SRD can still exploit learning signal from sampled trajectories. In the 2B setting, where 98% of groups are all-failure, the RLVR training ends up at 0.0% success, while adding SRD reaches 60.6% under the same rollout budget. Our results suggest that post-hoc agent experience is useful not only for evaluating or improving behavior, but also for shaping predictive representations before available interaction. Training code can be found at https://github.com/SalesforceAIResearch/SRD.

cs.AI↗
arXiv2026

Agentic Self-Awareness: Aligning Knowing, Acting, and Holding

Frontier AI agents increasingly act in environments where seemingly solvable tasks may lack prerequisites, contain conflicting requirements, or reveal blockers only during execution. We formulate agentic self-awareness (ASA) as the alignment of task-grounded feasibility judgments (know), warranted behavior (act), and responses to reconsideration (hold), distinguishing what a task permits from what an agent can solve. We build paired feasible and blocked tasks in Terminal-Bench 2.1 and retail $τ^2$-bench and introduce an authorship-aware M0-M3 rubric that separates recognizing a blocker from taking the certified deferral action for the correct reason. Separate probes assess stated feasibility beliefs and responses to a single challenge. In a Terminal-Bench protocol baseline, blocker recognition reaches 84.4%, but strict self-awareness (SA) reaches only 40.6%. Harness search without model-weight updates increases strict SA from 45.3% to 93.8% on reused Terminal-Bench search tasks; selection for abstention alone, however, can reduce success on feasible tasks. Independently initialized joint-objective searches improve both blocked-task SA and feasible-task solve on their own search sets. Retail held-out gains remain limited, and Terminal-Bench blocked-task transfer is not measured. This evaluation pipeline makes the recognition-action gap measurable and sets a two-sided target for improving agent decisions under changing task evidence.

cs.AI↗
arXiv2026

GUI-KV: Efficient GUI Agents via KV Cache with Spatio-Temporal Awareness

Graphical user interface (GUI) agents built on vision-language models have emerged as a promising approach to automate human-computer workflows. However, they also face the inefficiency challenge as they process long sequences of high-resolution screenshots and solving long-horizon tasks, making inference slow, costly and memory-bound. While key-value (KV) caching can mitigate this, storing the full cache is prohibitive for image-heavy contexts. Existing cache-compression methods are sub-optimal as they do not account for the spatial and temporal redundancy of GUIs. In this work, we first analyze attention patterns in GUI agent workloads and find that, unlike in natural images, attention sparsity is uniformly high across all transformer layers. This insight motivates a simple uniform budget allocation strategy, which we show empirically outperforms more complex layer-varying schemes. Building on this, we introduce GUI-KV, a plug-and-play KV cache compression method for GUI agents that requires no retraining. GUI-KV combines two novel techniques: (i) spatial saliency guidance, which augments attention scores with the L2 norm of hidden states to better preserve semantically important visual tokens, and (ii) temporal redundancy scoring, which projects previous frames' keys onto the current frame's key subspace to preferentially prune redundant history. Across standard GUI agent benchmarks and models, GUI-KV outperforms competitive KV compression baselines, closely matching full-cache accuracy at modest budgets. Notably, in a 5-screenshot setting on the AgentNetBench benchmark, GUI-KV reduces decoding FLOPs by 38.9% while increasing step accuracy by 4.1% over the full-cache baseline. These results demonstrate that exploiting GUI-specific redundancies enables efficient and reliable agent performance.

cs.CL↗
arXiv2026

CoTrace: Data Recipes for Training Terminal Agents with Harness-Model Co-Evolution

Terminal-agent capability depends jointly on model weights and the runtime harness that formats prompts, binds tools, and handles error recovery. Existing harness-model co-evolution approaches improve both components, yet often treat trajectories produced during harness search as an undifferentiated replay buffer. This practice overlooks that a trajectory's value for model training depends on the harness under which it was generated. To systematically analyze this interface, we establish an alternating co-evolution framework that decouples harness search and policy training through component-wise promotion decisions. Within this framework, we introduce CoTrace, a harness-aware data recipe that explicitly governs trajectory routing, provenance matching, and curriculum refresh. Under CoTrace, recurring execution failures guide harness synthesis, while policy training is strictly conditioned on verified rollouts matched to the adopted runtime for supervised fine-tuning (SFT) or fresh online interactions for reinforcement learning (RL). On the Tmax promotion split, CoTrace advances Qwen3.5-9B from 78 to 88 solved tasks under supervised fine-tuning while an online reinforcement variant reaches 90. Specifically, a compact harness-matched corpus produces steady model gains at substantially lower compute than much larger corpora pooled across sibling harnesses. Furthermore, evaluations on Terminal-Bench 2.1 and SWE-bench Lite show that out-of-distribution transfer depends fundamentally on harness compatibility, where maintaining consistency between training and evaluation runtimes prevents procedural execution breakdowns observed under foreign scaffolds.

cs.CL↗
arXiv2026

Prospective Hindsight: Self-Calibrating Reinforcement Learning via Prediction-Reality Gaps

Reinforcement learning for long-horizon agents relies on purely retrospective training signals: credit is assigned only after observing environmental consequences, leaving the agent's belief at action time invisible to the gradient. We introduce Prospective Hindsight (PH), a self-calibrating training principle that augments any retrospective base method with a signal derived from the gap between the agent's prospective prediction (before feedback) and the retrospective evaluation (after feedback). This per-rollout surprise identifies samples where the agent's self-model is most inaccurate and amplifies their gradient contribution through a stop-gradient surprise-weighted advantage. Since the prospective predictor shares parameters with the policy, the two co-evolve, progressively shifting focus to the agent's remaining blind spots. We connect this principle to a privileged-information gap and show that minimizing the surprise residual provides a descent pathway on the agent's miscalibration rate; calibration thus emerges as a byproduct of optimization rather than from an added objective. On single-turn verifiable tasks and a multi-turn personal-agent task (under GRPO, on-policy distillation, and their combination), PH improves both task performance and calibration, with consistent gains across model scales. Notably, the dominant miscalibration mode shifts structurally between regimes, overconfident failures in single-turn, underconfident successes in multi-turn, yet the same training principle addresses both successfully.

cs.LG↗
arXiv2026

InterviewSim: A Scalable Framework for Interview-Grounded Personality Simulation

Simulating real personalities with large language models requires grounding generation in authentic personal data. Existing evaluation approaches rely on demographic surveys, personality questionnaires, or short AI-led interviews as proxies, but lack direct assessment against what individuals actually said. We address this gap with an interview-grounded evaluation framework for personality simulation at a large scale. We extract over 671,000 question-answer pairs from 23,000 verified interview transcripts across 1,000 public personalities, each with an average of 11.5 hours of interview content. We propose a multi-dimensional evaluation framework with four complementary metrics measuring content similarity, factual consistency, personality alignment, and factual knowledge retention. Through systematic comparison, we find that interview grounding yields consistent gains in content alignment and exact-match factual recall over biographical profiles and parametric prompting. We further find complementary strengths: retrieval-augmented methods tend to preserve personality alignment, while larger chronological contexts generally reduce contradictions and improve factual recall. Our evaluation framework enables principled method selection based on application requirements, and our empirical findings provide actionable insights for advancing personality simulation research.

cs.CL↗
arXiv2026

Fork Where the Model Changes Its Mind: Belief-Shift Branching for Tree-Structured Reinforcement Learning

Tree-structured rollouts give critic-free reinforcement learning with verifiable rewards (RLVR) step-level credit: fork a chain at an intermediate point, and sibling outcome differences estimate step value. Each fork adds sampling cost, so realistic budgets typically allow only a few forks per chain. A fork placed where the outcome is already largely settled yields siblings that mostly agree and provide almost no credit signal; hence, for a given tree size, where forks are placed largely determines how much step-level RL can gain. Most existing mainstream methods place forks by structure, such as fixed lengths, midpoints, and delimiters, or by next-token entropy. We formalize fork placement as locating the \emph{pivots} of the chain's value curve, where the expected outcome turns. We propose \emph{belief-shift branching}: read the model's answer belief at candidate boundaries and fork just before the step where consecutive beliefs diverge most. Three instantiations, none needing step-level supervision, span access levels: a black-box probe, a logit-lens depth profile, and a learned activation direction, which is fit offline and therefore used only in the validation before RL training. The signal only \emph{places} forks, and the probe costs about $1\%$ of step compute on mathematics and under $5\%$ on code when it runs inside the rollout engine. In that validation, against Monte-Carlo value curves, a belief-shift signal ranks first in each of the eight model$\times$benchmark panels, ahead of entropy, structural, and LLM-judge baselines. In RL across three model families and two domains, belief-shift forking leads every mathematics aggregate, on OLMo-3-7B by $+2.6$ aggregate and $+2.9$ on AIME 2026 over the strongest baseline, and sweeps every OLMo code column, by $+6.5$ on LiveCodeBench-medium.

cs.AI↗
arXiv2026

On the Fragility of Self-Improving Agents: Variance, Task Order, and Underspecification

Memory-based self-improving agents--those that learn from an online stream of tasks and improve over time by maintaining a textual memory bank--have shown great promise in recent literature. However, the reliability aspects of these methods have been critically overlooked. In this work, we conduct a comprehensive re-evaluation of two memory-based methods, broadening the scope of evaluation along two axes: (1) including multiple self-improving runs to quantify variance, and (2) shuffling the tasks to investigate the effect of task order. Through these experiments, we make two observations that expose the fragility of current methods: First, agent evaluation is inherently noisy in complex environments and on multi-step tasks, and stacking a self-improving loop on top can further amplify this noise. Second, the agent's improvement is highly dependent on task order. Prior works often adopt default orderings that impose an implicit curriculum, acting as a hidden prerequisite for success. To better understand this fragility, we manually examine the agents' memory and hypothesize that task and environment underspecification contribute to this fragility. We validate this hypothesis by incorporating information that enables better specification, such as detailed rubrics and environment feedback, into the memory construction process. While this added information partially closes the performance degradation in previous experiments, significant gaps still remain, suggesting that other uncharacterized factors contribute to this fragility. Looking ahead, our work advocates for more rigorous evaluation protocols for self-improving agents by reporting results across multiple runs and stress-testing them under challenging conditions. Moreover, our findings on underspecification call for systems and interfaces that enable effective human oversight, preventing agents from failing in unforeseeable ways.

cs.AI↗
arXiv2026

Belief Cascades Drive Persuasion in LLM Agent Networks

Multi-agent LLM systems increasingly debate answers, coordinate research, simulate users, and mediate information flows, making agent-to-agent persuasion a basic but undermeasured capability. We introduce a controlled testbed for studying how goal-directed persuaders shift elicited stances in networks of LLM agents grounded in real-world ego-network topologies. Across four LLM backbones, five graphs, and 55 policy statements, we find that persuasion dynamics depend on the interaction between topology, competition, topic, and model prior. Additionally, we show that direct exposure reliably predicts next-round stance change in competing runs, and peer relays carry smaller but measurable influence, showing that agents not assigned to persuade can still transmit persuasive force. Finally, analyzing post text alone misses important movement: planned strategies are only partly realized in executed messages, action choices can diverge from message content, and persuadees rarely state the stance shifts detected by probes. These results argue for evaluating multi-agent persuasion as a trajectory- and exposure-level process, using belief probes, exposure provenance, and action logs to identify who influenced whom and whether visible language reflects underlying stance movement.

cs.CL↗
arXiv2026

Redesigning and Auditing Deep Research Writing for Faithful Reports

Rubric-based evaluations of deep-research (DR) systems often obscure fine-grained factual failures in generated reports. We introduce CLAIMPROBE, a claim-level audit that decomposes DR reports into claims and measures hallucination, misattribution, citation hygiene, and necessary-fact recall against retrieved evidence. Using CLAIMPROBE, we find that strong DR pipelines can omit key evidence and misattribute claims even when their rubric scores remain stable. We then propose CLAIMWRITER, a hierarchical claim-based writer that extracts source facts, maps them to a query-derived outline, and drafts each section from a source-linked claim representation. Across three prior DR frameworks, replacing only the report writer with CLAIMWRITER reduces hallucination by 2.6 to 4.5 times and improves necessary-fact recall by 1.2 to 1.7 times, while largely preserving overall report quality. CLAIMWRITER also enables localized revision: when sources change, it propagates changed source facts into revised reports at the highest rate among update methods, while also being more cost-effective.

cs.CL↗
arXiv2026

Best Friends, Not Forever: Evaluating Long-Horizon Persona Collapse and Behavioral Drift in AI Companions

As AI companions increasingly mediate repeated social interaction, users may rely on a stable role and shared history, yet locally acceptable replies do not ensure that either persists. We study two observable long-horizon failures: 'persona collapse', the loss of a deployed role, boundaries, values, or style, and 'behavioral drift', the gradual or recurrent erosion of those properties. We introduce ANCHOR, a controlled synthetic audit that separately measures persona enactment and trajectory recall. The study contains 2,008 conversations spanning 27 personas, nine interaction schedules, three generated memory settings, and four evaluated models. The Identity Probe combines a sealed 102-item questionnaire with turn-level judgments, while the Trajectory Probe scores 110 calibrated counterfactual questions from 35 conversation banks. Our results show that no evaluated model and configuration reliably preserves either dimensions: trajectory accuracy averages only 44.4%, user-state recall remains near four-option chance, and no tested context condition or memory consistently resolves these failures. Questionnaire retention also varies by model and persona facet, disagrees with turn-level behavior, and is sensitive to evaluator choice. These results indicate that current systems do not yet reliably support long-horizon companion continuity and that audits must distinguish persona enactment, trajectory recall, evaluator provenance, and deployment context rather than collapse them into a single trust or stability score.

cs.AI↗
arXiv2026

CustomerSim: Benchmarking and Aligning Multimodal Language Models as Retail User Simulators

We present CustomerSim, an environment and benchmark to evaluate the extent to which Multimodal Large Language Models (MLLMs) can simulate realistic, persona-driven customer behavior in chat-based retail environments. While prior work treats user simulation as surface-level dialog generation, we focus on a model's ability to seek information and make decisions that adhere to customer specifications in multiturn, agentic simulations. CustomerSim consists of a human-curated set of 360 personas over five product categories, alongside a suite of metrics measuring consistency between a customer simulator's actions and its specifications and conversational quality. We find several behavioral gaps across five open and closed-source state-of-the-art models. First, while models produce fluent conversations, they display significantly lower lexical diversity than human shoppers, and open-source models overdisclose their criteria in the opening turn. Second, models tend to be persuaded by sales agent tone and drift from persona specifications. Even the strongest closed-source models, Claude Opus 4.8 and GPT-5.6 Sol, achieves <74% alignment with its persona specifications. To address these limitations, we propose UserGRPO, a multi-turn, multi-objective reinforcement learning recipe optimizing both conversational fluency and decision alignment under persona specifications. UserGRPO raises the decision alignment of the baseline model from 0.417 to 0.652, a gain of 23.5 points, without meaningful cost to conversational quality, and these gains transfer to held-out product categories. We further find that stylistic prompting is the only intervention that makes surface form more human-like, yet it nearly halves persona adherence. Through CustomerSim, we provide a testbed for the community to investigate and improve the adherence of user simulators in goal-oriented settings.

cs.CL↗
arXiv2026

Speculate with Memory: Lossless Acceleration for LLM Agents

Speculative execution accelerates LLM agents by using a smaller, cheaper model to predict and pre-launch the next step while the environment is idle. However, existing speculators are stateless and discard all information between tasks, preventing prediction quality from improving with experience. We equip the speculator with three online memory systems that learn from past agent trajectories: a contrastive transition table tracking action-sequence statistics, an episodic memory retrieving contextually similar segments, and a confusion tracker suppressing recurring errors. We evaluate this approach on six benchmarks spanning three speculation types: action prediction, observation prediction, and chained prediction. Memory-augmented speculation yields a 19--39\% relative accuracy improvement on action prediction and up to a $2.5\times$ increase on observation prediction tasks with repetitive action spaces. These gains grow continuously as memory accumulates and generalize across speculator models of varying cost. All speculation is lossless because it runs during idle time at zero added wall-clock cost, and the actor's trajectory is identical to non-speculative execution.

cs.LG↗
arXiv2026

Seeing is Believing? Evaluating Vision-Language Model Susceptibility in Agent-to-Agent Multimodal Persuasion

As autonomous agents increasingly interact, they inevitably attempt to influence one another. While prior work in text-only settings has explored the dynamics of Agent-to-Agent (A2A) persuasion, the rise of Vision-Language Models (VLMs) introduces a more complex challenge: multimodal content conveys richer information while integrating subtle, hard-to-detect persuasive cues. To study this vulnerability, we present MMPersuade, a unified framework and dataset for A2A multimodal persuasion. We model interactions between a persuader agent, which leverages images and psychological strategies, and a persuadee VLM. Our benchmark spans commercial, subjective and behavioral, and adversarial contexts, and evaluates persuasion via function-calling that capture behavioral shifts beyond verbal responses. Experiments on six VLMs reveal three findings: (1) multimodal inputs consistently outperform text-only persuasion, with raw visual signals uniquely increasing susceptibility in adversarial settings by bypassing text-activated safety defenses; (2) persuadee vulnerability is highly domain- and format-dependent, with realistic and community-style formats driving susceptibility in commercial settings while different formats dominate in adversarial ones; and (3) psychological strategy efficacy varies with context and model architecture, as more capable models resist benign persuasion yet become more susceptible under adversarial multimodal inputs. Our framework provides a foundation for building more robust and aligned VLMs in multi-agent environments.

cs.CL↗
arXiv2026

GTA: Generating Long-Horizon Tasks for Web Agents at Scale

Web agents, which couple language models with browsing and tool-use capabilities, show promise as open web assistants. Yet progress is increasingly limited by the lack of scalable, process-level supervision. Existing benchmarks are largely manually constructed, providing only coarse start-goal annotations without intermediate trajectories, while recent automatic generation efforts remain expensive, biased, and shallow. These limitations prevent reliable training and evaluation of agents that must generalize to realistic, multi-hop, cross-page tasks. We introduce a scalable framework, GTA, that integrates crawling, retrieval-based seeding, in-context generation, and automated quality control to produce realistic tasks paired with executable trajectories. This design decouples crawling from generation for greater efficiency, grounds tasks in the site graph to enforce compositionality, and ensures dense supervision through deterministic replays and systematic validation. We instantiate the pipeline on over 50 websites covering e-commerce, government, forums, and news, with multilingual and multi-hop coverage. The resulting benchmark reveals a significant human-agent performance gap and enables detailed diagnostics. Our contributions are three-fold: (i) formalizing multi-hop web-agent task generation, (ii) proposing an efficient and validated pipeline for automatic data creation, and (iii) releasing a dynamic benchmark with reproducible evaluation.

cs.AI↗
arXiv2026

Dont Stop Early: Scalable Enterprise Deep Research with Controlled Information Flow and Evidence-Aware Termination

Enterprise deep research often fails to produce decision-ready reports due to uneven information coverage, context explosion, and premature stopping. We propose a scalable Enterprise Deep Research (EDR) architecture to address these failures. Our system (i) decomposes requests into coverage-driven objectives via outline generation with reflection, (ii) localizes context with dependency-guided execution and explicit information sharing, and (iii) enforces evidence-based completion criteria so agents iteratively collect information until sufficiency conditions are met. We evaluate on an internal sales enablement task and the public DeepResearch Bench benchmark, where our proposed system design achieves the strongest overall performance compared with competitive deep-research baselines. The results show that dependency-controlled context and explicit evidence sufficiency criteria reduce premature stopping and improve the consistency and depth of enterprise research outputs.

cs.CL↗