arXiv subjects

Zhangkai Wu

Publications and source records attributed to Zhangkai Wu.

At least 19 recordsLinked to original sources

arXiv2026

A State Change Is Local, Its Consequences Are Not: Dependency-Aware State Revision for Long-Term Agent Memory

Long-lived agents rely on persistent memory of users, tasks and decisions, and this memory must stay consistent with a world that keeps changing. When an upstream fact changes, however, memory systems can revise the contradicted memory itself while the beliefs, plans and actions derived from it remain accepted, even though they have lost their support. We therefore formulate dependency-aware state revision, which asks not only what changed but also which dependent states should no longer be trusted. To address it, StateGraph maintains an evolving state dependency graph whose edges are recovered from interaction histories. Its core design is counterfactual verification: each candidate edge is tested by asking whether the dependent state would survive the loss of its premise, which separates support dependence from semantic relatedness, and revision then invalidates a dependent state only when a strict dependency fails and no independent support path survives. To test this design, we introduce StateChangeBench, a controlled benchmark of 50 author-constructed cases annotated with cascade depth, hard negatives and action consequences. On it, StateGraph's Impact-F1 margin over a CUPMem reimplementation is 6.9 percentage points at depth 1 and 26.0 at depth 3+. On public long-term memory benchmarks, its point estimates are comparable to or higher than those of the evaluated baselines on most datasets. These results provide initial evidence that explicit dependency tracking can improve selective memory revision under changing conditions.

cs.CL↗

arXiv2026

CCD: Context-Calibrated Open-Entity Out-of-Distribution Detection in In-Context Tabular Foundation Models

In-context tabular foundation models (TFMs), such as TabPFN, predict from a labelled context without updating their parameters and are increasingly used for out-of-distribution (OOD) detection in Web applications, whose test rows come from entities absent from that context. Such applications need open-entity OOD detection: a new entity must not be mistaken for an unknown class, and an unknown class must still be caught. Existing detectors fail in this setting, and an entity-split gap hides the failure: random splits also place the test entities in the context. The failure arises because every TFM score reflects how unfamiliar a row is to the context. This produces an entity-over-class bias: known-class rows of new entities look more OOD than familiar ones, sometimes more than unknown classes. Specifically, the two OOD scores most widely used with in-context TFMs face an entity-class trade-off: the TFM's predictive confidence misses unknown classes, whereas the distance of a test row to the context flags new entities. To address this problem, we hold a group of entities out of the context and score them with the remaining context, which turns them into new entities of known classes and tells us how such rows should score. To measure detection on new entities, we build NEO-Tab (New-Entity OOD detection on Tables), 14 grouped and temporal tables on which all test rows come from new entities. We then propose Context-Calibrated Detection (CCD), which uses these group-out swaps on in-distribution data to calibrate three distances and the confidence against new-entity references. CCD weights them by their sensitivity and sharpness, keeps the model frozen and uses no OOD data. On NEO-Tab, CCD outperforms 16 existing detectors on every metric we report, and its lead holds on three TFM backbones.

cs.IR↗

arXiv2026

Recall or Relearn? A Mixture of In-Context Experts for Recurring Concept Drift

Tabular foundation models (TFMs) are being applied to Web data streams such as network intrusion detection, where they adapt by updating the labelled rows of their context without retraining. When a past concept returns after leaving the context, a TFM faces a recall-or-relearn decision between restoring stored rows of the concept and relearning it from new labels. We propose MICE (Mixture of In-Context Experts), a training-free learner that cuts the stream into segments at detected changes. It merges a closed segment into a stored context when the TFM's estimated additional error on the segment is within a tolerance, under a capacity limit on stored contexts. A fast level weights the stored contexts and first-in-first-out (FIFO) windows by the log-loss of their predictions on the last labelled batch. We show that, with pure stored contexts and a margin condition, this fast level gains over relearning, in expected accuracy, the normalised positive area between the stored context's accuracy and the TFM's learning curve, up to a first-batch term. A slow level mixes the fast level's prediction with that of the FIFO window of the last labelled rows and bounds the discounted half-Brier regret of MICE against this window at every time. Learning curves measured beforehand predict the ordering of the observed recall gains. MICE outperforms reset, which empties the context at detected drifts, on grid streams with random boundary phases. Merging keeps fewer experts than pools that store every segment, and with fixed segmentation a time-matched snapshot pool matches MICE on the shifted grid. On real streams, MICE has the highest mean accuracy of all policies we compare.

cs.LG↗

arXiv2026

From Faulty Memories to Corrected Actions: Dependency-Guided Rollback Repair for Memory-Augmented Agents

Persistent memory lets language model agents reuse information across sessions, but it also makes errors durable: a poisoned, stale, or misattributed record can alter reasoning, tool use, answers, and subsequent memory writes. Existing defenses mainly detect or delete suspicious memories, or revise the current response. Deleting the source leaves already propagated claims, actions, and derived memories active, whereas resetting the store or replaying the full trace destroys benign state and repeats unnecessary computation. We therefore formulate post-failure memory recovery: given a failed execution and diagnosed faulty memories, regenerate the answer and clean up affected persistent state while retaining unaffected work. Our dependency-guided rollback repair builds a typed memory-to-action graph from runtime provenance, traces explicit downstream dependencies, preserves candidates with independent trusted support, deactivates unsupported memory state, and selectively replays only answer-relevant affected computation. We evaluate this approach on a 150-case controlled benchmark spanning three tool-use domains and four memory failure types, and on a 50-case trajectory-derived stress test adapted from LongMemEval-V2. On the controlled benchmark, it achieves 85.3% immediate recovery versus 77.3% for LLM-judge repair, the strongest baseline, while preserving all benign memories and using 41.8% fewer LLM calls. Selective replay, however, leaves higher recurrence than LLM-judge repair (26.6% vs. 12.1%), exposing a trade-off between targeted repair and broader recomputation. On the adapted subset, it reaches 68.0% recovery versus 54.0% for the next best method, with 5.9% recurrence. Code and data are available at https://github.com/xiaoqi-7/memory-rollback.

cs.AI↗

arXiv2026

Memories That Outlive Their Reasons: Dependency-Tethered Memory for Tool-Using LLM Agents

Agents now act on the Web, publishing packages, models and datasets through registry APIs, and remembering past verifications spares them the work. But a verdict rests on premises. When a signing key is revoked, everything built on it stops passing, yet the advisory names only the key, and the old verdicts of its dependents may never share a retrieval window with it. This non-locality of invalidation turns memory into a liability: in long simulated chats, an agent reading a memory of atomic facts publishes 62% of the objects whose failure lies upstream, each backed by a record that was true when written. Yet derivation is local even when invalidation is not. Tether records, at write time, the exact versions each verdict derives from, and at read time withdraws any verdict whose support has failed anywhere in the store; the agent reads the same fields as before. On Fallout, a generator of versioned worlds with hypothesis-blind simulated histories shared by every memory, Tether publishes 0.3% of upstream-failing objects across three domains when relations are stated with their verdicts, with 37% fewer verification calls than checking everything. A-Mem and Graphiti re-verify most candidates (0.83--0.86 checks per candidate), which limits their savings, yet publish 9% of upstream-failing objects; Mem0 relies on memory more and publishes 42%. Flushing every verdict on any failure is as safe as Tether but costs 32% more checks. The gain over fact memory holds with three open-weight backbones, a whole-store window and real Go advisories. These gains depend on capturing dependencies: when relations arrive in sessions apart from their verdicts, Tether's downstream unsafe publication rate rises to 9.4--22.1%. It also degrades with weaker writers that miss dependencies or invalidation events.

cs.CL↗

arXiv2026

When Is a Memory Safe Enough to Act On? Action-Conditioned Verification for Long-Lived Agents

Long-lived agents act on memory carried across sessions, and many of their actions are irreversible, so a stale memory can turn directly into harm. Existing work improves the memory itself --- retrieving, refreshing, or reconciling entries --- yet whether an entry must be re-checked before acting is not a property of the entry: the same remembered address can be used unchecked for a routine shipment but should be re-confirmed before a valuable one. We therefore study \emph{action-conditioned memory verification}: given a candidate action, decide which of its memory premises to re-verify and when to proceed or abstain, balancing execution risk, abstention, and verification cost. We formulate this as a sequential stopping problem, in which a verification is worth its cost only if it can change what the agent does next; under suitable verifier reliability and cost conditions, increasing action consequence can move the same memory through proceed, verify, and direct-abstention regimes. Building on this formulation, \textsc{AMV-Guard} is a practical verification gate that extracts the premises a candidate action rests on, calibrates beliefs about whether they are wrong, and selects checks by bounded value of information (VOI). On ToolSandbox, relative to a static policy given the same inputs, it lowers mean verification cost by 33.9\% while keeping 100\% safe-proceed recall and 0\% Gate harmful, and raises Proceed coverage from 38.83\% to 44.17\%; in a gate-core transfer to AppWorld, which carries over the stopping/VOI core but not the full structured pipeline, it lowers verification cost by 21.3\% at a small cost in safe-proceed recall and task success. These results suggest that verification effort should be set by what an agent is about to do, not only by how reliable a memory appears.

cs.MA↗

arXiv2026

From Agent Traces to Trust: A Survey of Evidence Tracing and Execution Provenance in LLM Agents

Large language model (LLM)-based agents are moving beyond text generation to planning, tool use, retrieval, memory access, environmental interaction, and multi-agent collaboration. Because their outputs depend on information and execution steps across an agent run, their execution processes are difficult to verify, debug, and audit. Even a correct final answer does not reveal which evidence supported its claims, whether tool calls were justified, how memory shaped decisions, or where execution failures originated. Relevant research addresses these questions across evidence attribution, tool-use safety, agent memory, and observability, but these areas emphasize different parts of the execution process. This survey brings them into a common framework through execution provenance, a typed graph of an agent run, and evidence tracing, its projection onto evidence-support relations, distinguishing evidence that supports a conclusion from execution that influences it. We introduce a six-dimensional taxonomy covering trace sources, evidence and execution units, provenance relations, tracing granularity and timing, representation forms, and trust functions, complemented by operational definitions of nine provenance relations and two cross-cutting attributes, provenance origin and trace assurance. Applying this framework to the surveyed systems and benchmarks reveals three gaps: limited integration of fine-grained provenance with multiple trust functions, no benchmark in our catalog that evaluates the complete provenance chain, and little reporting of tamper-evident or attested traces. These findings motivate research on unified trace schemas, semantic provenance, memory and multi-agent lineage, provenance-aware safety and recovery, realistic benchmarks, trace assurance, and privacy-aware auditing. Resources are available at https://github.com/xiaoqi-7/Agent-Tracing-Survey.

cs.CR↗

arXiv2026

IndustrialVLA-Bench: A Traceable Multi-Axis Evaluation of Open Robot Policy Models

Open robot policies increasingly follow two paradigms: vision-language-action models (VLAs) directly map observations and instructions to actions, whereas world-action models (WAMs) incorporate learned video or world dynamics into policy learning or action generation. Although both target the same manipulation tasks and represent alternative design choices, they are commonly reported under different evaluation protocols, leaving their capability, robustness, language sensitivity, and deployment-cost trade-offs unclear. We present IndustrialVLA-Bench, an evidence-aware evaluation of six released VLA and WAM systems under a unified reporting schema. It separately evaluates clean capability on LIBERO, non-language robustness on LIBERO-Plus, instruction sensitivity on LIBERO-Para, and observed execution cost. Reported task scores aggregate three complete evaluations with distinct random seeds under a fixed checkpoint and inference configuration. Across all six systems, clean LIBERO averages differ by only 1.58 points, whereas robustness and paraphrase summaries span 14.62 and 31.08 points. Restricting every comparison to the three protocol-faithful systems preserves the effect (1.36, 14.62 and 23.10 points), so the diagnostic separation reported here does not depend on the weaker evidence tiers. We additionally report observed inference latency, peak memory, runtime mode, and an evidence status for every system. Protocol-faithful, near-reproduction, and pending-verification entries remain visibly separated; only protocol-faithful entries support strict comparisons. Rather than claiming universal superiority of either paradigm, IndustrialVLA-Bench provides traceable evidence for comparing released robot policies on shared practical criteria. Code and evaluation records are available at https://github.com/xiaoqi-7/IndustrialVLA-Bench.

cs.RO↗

arXiv2026

PACMI: Provenance-Aware Cascading Memory Invalidation for Long-Term LLM Agents

LLM agents rely on long-term memory to retain and reuse information when performing tasks over long horizons. Existing methods provide limited support for handling memories that become outdated as new observations or domain evidence arrive. Such outdated memories may remain semantically relevant, continue to affect dependent records, and retain value as historical evidence. This calls for two capabilities: dependency tracking to identify downstream effects and historical preservation to retain useful past records. We propose Provenance-Aware Cascading Memory Invalidation (PACMI), a framework that represents memories and new evidence in a provenance graph with typed dependency edges. PACMI assigns records to a four-state validity lattice, propagates validity changes to dependent memories, and uses the resulting states for retrieval and stale-premise detection. We also introduce a diagnostic benchmark with 100 cases and 300 queries across five domains. The evaluation separates node, context-, and answer-level performance. PACMI achieves the highest final-answer accuracy on this benchmark, and its paired difference from the strongest baseline is significant under an exact McNemar test. The premise checker achieves perfect precision, recall, and F 1 on the controlled query distribution. Cascading propagation primarily improves memorystate correctness: removing it increases final-answer errors from 3 to 11, but the paired difference does not reach the 0.05 significance threshold. Code and data will be made publicly available.

cs.LG↗

arXiv2026

Heterogeneous Dependency Graph-Guided Attentionfor Patent Representation Learning

Pre-trained language models advance patent classification and retrieval by encoding claims as flat token sequences, but they overlook the dependency hierarchy among claims. Incorporating this hierarchy into self-attention poses two challenges. First, claim dependencies include relation types with different levels of reliability, so treating them uniformly may allow noisy technical relations to interfere with more reliable legal citations. Second, claim dependencies are defined at the claim level, whereas Transformer attention operates over tokens, making direct structural injection nontrivial. We propose the Patent Heterogeneous Attention-Guided Graph Encoder (PHAGE), which constructs a typed claim graph that distinguishes legal citations from technical relations. PHAGE projects this claim-level topology into token-level attention through a connectivity mask and learnable relation-aware biases, and fine-tunes the encoder using a dual-granularity contrastive objective that combines inter-patent taxonomy with intra-patent topology. At inference, the graph-specific attention components are removed, allowing representations to be generated through a standard encoder forward pass without CDG construction. Experiments on patent classification, retrieval, and clustering show that PHAGE consistently outperforms domain-adapted and citation-aware baselines, demonstrating the value of claim-level structural guidance for graph-free patent representation learning.

cs.CL↗

arXiv2026

GAPL: Grounded Action-effect Policy Learning for LLM-Based Trajectory Planning

Trajectory planning for autonomous driving requires both high-level reasoning and precise low-level control. Large Language Models (LLMs) offer semantic-rich planning capabilities, however, their application is limited by hallucinated reasoning, poor grounding in environment dynamics, and limited numerical precision in control. We propose GAPL (Grounded Action-effect Policy Learning), a unified framework that integrates LLM-based effect estimation, simulation-based effect grounding, and policy optimization into a closed-loop system. GAPL consists of three modules: (1) an LLM-based Effect Evaluator for structured multi-dimensional action-effect estimation; (2) a Simulation-based Effect Grounder that predicts dynamics-consistent effects from simulator rollouts; and (3) an Effect-Aware Decision Maker that grounds LLM effect estimates against simulation via a distiller to guide Proximal Policy Optimization (PPO)-based policy learning. Experiments on four Highway-env scenarios demonstrate that GAPL consistently outperforms baselines, achieving average reductions of {0.76, 0.86, 2.00} in collision rate, average displacement error (ADE), and final displacement error (FDE), and an average reward gain of 1.44.

cs.RO↗

arXiv2026

MAP-Graph: Provenance-Aware Shared Memory for Multi-Agent Workflows

Shared memory helps language-model agents reuse information across long workflows, yet relevant evidence may not be admissible for a particular agent or action. Because restrictions propagate through derivations, summaries can conceal private, poisoned, untrusted, or revoked sources, enabling unauthorized reads or unsafe actions. Existing approaches provide semantic retrieval, scoped access, or lineage tracking, but do not clearly separate hard authorization from graded trust or adapt evidence requirements to action risk. We introduce MAP-Graph, a provenance-aware memory layer that represents agents, sources, memories, claims, and actions in a typed execution graph. It traces ancestry, excludes permission-ineligible records, reranks eligible memories by semantic similarity and multiplicative path trust, and applies a risk-sensitive gate before action execution while retaining affected lineage for audit. On a controlled benchmark of 2,700 synthetic tasks per method across three domains, MAP-Graph achieves 94.96\% overall task success, 72.70\% exact decision accuracy, and 90.22\% in the clean setting, where success requires a correct \textsc{Allow} rather than a safe intervention. Ablations isolate the roles of permission filtering, path trust, and action gating, while transfer tests with two additional backbones preserve the exact-decision and access-control advantages. These results support provenance as an operational control signal, rather than only post-hoc audit metadata, within the evaluated setting.

cs.AI↗

arXiv2026

Pointer-Augmented Autoregressive Generation of Patent Claims with Joint Topology and Content Decoding

Autoregressive decoders emit flat token sequences and cannot enforce hierarchical constraints across output segments, a limitation that becomes acute in patent claim generation, where a claim set forms a dependency forest whose scope must narrow monotonically with depth. Topology and content are mutually dependent: a dependent claim's wording must reflect its parent's scope, yet the parent must be chosen before that wording exists, so neither post-hoc parsing nor grammar-constrained decoding suffices. We propose SPG (Structure-aware Patent Generation), which predicts topology inside the autoregressive pass. A pointer head selects each dependent claim's parent, and its gradients, together with a depth-adaptive scope regularizer, reshape the shared decoder's representations during training. A second stage then applies a violation-weighted preference objective over self-generated deficient candidates, supplying the negative signal that granted-patent corpora lack. On HUPD-DCG, SPG on Llama-3-8B-Instruct recovers 79.0\% of gold parent links, a quantity its training reward never supervises, and raises antecedent consistency from 0.292 to 0.478 over a supervised baseline of equal scale, with expert evaluation corroborating these gains.

cs.CL↗

arXiv2026

Reliable AI Needs to Externalize Implicit Knowledge: A Human-AI Collaboration Perspective

This position paper argues that reliable AI requires infrastructure for human validation of implicit knowledge. AI learns from both explicit knowledge (papers, documentation, structured databases) and implicit knowledge (reasoning patterns, debugging processes, intermediate steps). Implicit knowledge remains unexternalized because documentation cost exceeds perceived value -- yet AI learns from it indiscriminately, acquiring both beneficial patterns and harmful biases. Current reliability methods can only verify explicit knowledge against sources, creating a fundamental gap: the most valuable AI capabilities (reasoning, judgment, intuition) are precisely those we cannot verify. We propose Knowledge Objects (KOs) -- structured artifacts that externalize implicit knowledge into forms humans can inspect, verify, and endorse. KOs transform verification economics: what was previously too costly to verify becomes feasible, enabling accumulated human validation to improve reliability over time.

cs.AI↗

arXiv2026

DUET: Unified Dual-Space Emotion Control for Diffusion and Flow-Matching Driven Text-to-Speech

Diffusion and flow-matching based text-to-speech (TTS) models excel in naturalness but often lack explicit emotion control, as emotional signals remain entangled with speaker identity. We discover that emotion embedding emerges as a linearly decodable direction of frozen hidden states, nearly orthogonal to the direction embedding speaker identity. This inspires a plug-and-play framework DUET for emotion control over pretrained diffusion and flow-matching based TTS models. During generation, DUET unifies dual-space control to achieve fine-grained emotion intervention in a single per-step update: hidden space steering shifts generation along the target emotion direction, while mel-space guidance refines spectral details through gradients backpropagated from a differentiable vocoder. We validate DUET on five architecturally diverse pretrained TTS backbones across three datasets, where it outperforms 10 supervised state-of-the-art emotional TTS baselines across paradigms and achieves the highest human-rated emotion appropriateness. To further showcase its qualitative behavior, we deploy DUET on an Ameca humanoid robot, where it produces richly expressive emotional speech on the humanoid, demonstrating the strong potential for plug-and-play affective interaction for embodied agents.

cs.SD↗

arXiv2026

Learning Physiology-Informed Vocal Spectrotemporal Representations for Speech Emotion Recognition

Speech emotion recognition (SER) is essential for humanoid robot tasks such as social robotic interactions and robotic psychological diagnosis, where interpretable and efficient models are critical for safety and performance. Existing deep models trained on large datasets remain largely uninterpretable, often insufficiently modeling underlying emotional acoustic signals and failing to capture and analyze the core physiology of emotional vocal behaviors. Physiological research on human voices shows that the dynamics of vocal amplitude and phase correlate with emotions through the vocal tract filter and the glottal source. However, most existing deep models solely involve amplitude but fail to couple the physiological features of and between amplitude and phase. Here, we propose PhysioSER, a physiology-informed vocal spectrotemporal representation learning method, to address these issues with a compact, plug-and-play design. PhysioSER constructs amplitude and phase views informed by voice anatomy and physiology (VAP) to complement SSL models for SER. This VAP-informed framework incorporates two parallel workflows: a vocal feature representation branch to decompose vocal signals based on VAP, embed them into a quaternion field, and use Hamilton-structured quaternion convolutions for modeling their dynamic interactions; and a latent representation branch based on a frozen SSL backbone. Then, utterance-level features from both workflows are aligned by a Contrastive Projection and Alignment framework, followed by a shallow attention fusion head for SER classification. PhysioSER is shown to be interpretable and efficient for SER through extensive evaluations across 14 datasets, 10 languages, and 6 backbones, and its practical efficacy is validated by real-time deployment on a humanoid robotic platform.

cs.SD↗

arXiv2025

FAME: Fairness-aware Attention-modulated Video Editing

Training-free video editing (VE) models tend to fall back on gender stereotypes when rendering profession-related prompts. We propose \textbf{FAME} for \textit{Fairness-aware Attention-modulated Video Editing} that mitigates profession-related gender biases while preserving prompt alignment and temporal consistency for coherent VE. We derive fairness embeddings from existing minority representations by softly injecting debiasing tokens into the text encoder. Simultaneously, FAME integrates fairness modulation into both temporal self attention and prompt-to-region cross attention to mitigate the motion corruption and temporal inconsistency caused by directly introducing fairness cues. For temporal self attention, FAME introduces a region constrained attention mask combined with time decay weighting, which enhances intra-region coherence while suppressing irrelevant inter-region interactions. For cross attention, it reweights tokens to region matching scores by incorporating fairness sensitive similarity masks derived from debiasing prompt embeddings. Together, these modulations keep fairness-sensitive semantics tied to the right visual regions and prevent temporal drift across frames. Extensive experiments on new VE fairness-oriented benchmark \textit{FairVE} demonstrate that FAME achieves stronger fairness alignment and semantic fidelity, surpassing existing VE baselines.

cs.CV↗

arXiv2025

VALA: Learning Latent Anchors for Training-Free and Temporally Consistent

Recent advances in training-free video editing have enabled lightweight and precise cross-frame generation by leveraging pre-trained text-to-image diffusion models. However, existing methods often rely on heuristic frame selection to maintain temporal consistency during DDIM inversion, which introduces manual bias and reduces the scalability of end-to-end inference. In this paper, we propose~\textbf{VALA} (\textbf{V}ariational \textbf{A}lignment for \textbf{L}atent \textbf{A}nchors), a variational alignment module that adaptively selects key frames and compresses their latent features into semantic anchors for consistent video editing. To learn meaningful assignments, VALA propose a variational framework with a contrastive learning objective. Therefore, it can transform cross-frame latent representations into compressed latent anchors that preserve both content and temporal coherence. Our method can be fully integrated into training-free text-to-image based video editing models. Extensive experiments on real-world video editing benchmarks show that VALA achieves state-of-the-art performance in inversion fidelity, editing quality, and temporal consistency, while offering improved efficiency over prior methods.

cs.CV↗