World models are transitioning from passive visual generators to foundational, operational infrastructure for Physical AI: they must natively acquire world knowledge from heterogeneous experience, maintain persistent states over long horizons, and execute efficiently within real deployment constraints. We introduce Kairos, a native world model stack designed around these requirements. (1) Kairos learns the world by pioneering a Native Pre-training Paradigm governed by a Cross-Embodiment Data Curriculum, which organizes open-world videos, human behavioral data, and robot interactions into a progressive developmental pathway. (2) Kairos maintains the world by unified world understanding, generation, and prediction within a Native Unified Architecture equipped with Hybrid Linear Temporal Attention, where sliding-window attention captures local dynamics, dilated sliding windows capture mid-range dependencies, and gated linear attention maintains persistent global memory. We establish formal theoretical bounds demonstrating that this temporal factorization strictly limits error accumulation, mathematically guaranteeing state propagation across extended horizons. (3) Kairos runs the world by incorporating a Deployment-Aware System Co-Design to support low-latency rollout generation on server and consumer-grade hardware for real-world observation-action-feedback loops. Experiments on embodied world-model, long-horizon, and action-policy benchmarks show that Kairos achieves top level performance while offering a strong efficiency-capability trade-off. Together, these results position Kairos as a cohesive operational foundation for future self-evolving physical intelligence.

Consistent video generation under editing operations requires persistence: when edits modify scene appearance or layout, subsequent generations should remain coherent across time and viewpoints. However, existing memory designs struggle to maintain long-term consistency after such modifications, as stored contexts may become outdated or invalid. To address this, we propose PermaVid, a novel framework built upon a multi-modal context memory that disentangles spatial context into semantic appearance and geometric structure, together with an edit-aware memory update and retrieval strategy that keeps memory evolution aligned with subsequent observations. Specifically, we develop two complementary memory banks: an RGB context memory that captures appearance-aware observations while implicitly encoding geometry, and a depth context memory that preserves geometry-only structure disentangled from semantics. Building on this design, we introduce a memory-guided video generation model that performs multi-modal feature fusion under reference conditions drawn from mixed-modality memory contexts. Experiments demonstrate that our method maintains strong long-term semantic and structural consistency after edits, significantly outperforming state-of-the-art methods.

Evaluating representation similarity is fundamental to representation learning. However, existing metrics suffer from significant limitations: they lack interpretability due to shifting baselines, lack robustness to outliers, and are computationally intractable for large datasets, forcing reliance on heuristic approximations. To address this, we develop an ordinal-similarity framework, instantiated by the Triplet (TSI) and Quadruplet (QSI) Similarity Indices, which measure alignment by quantifying the consistency of ordinal relationships. We theoretically demonstrate this formulation is inherently interpretable, robust to outliers, and computationally efficient. Finally, we establish a formal equivalence between TSI and local neighborhood alignment, measured by Mutual Nearest Neighbors. Empirically, we validate these properties and show that ordinal similarity offers a scalable approach to measuring alignment, enabling practitioners to better understand and design representations.

Predictive modeling for clinical tabular data is central to clinical decision support and therefore requires not only strong predictive performance but also transparent decision logic. Although deep learning and tree-based ensemble methods can achieve high accuracy, their black-box nature remains a major obstacle to clinical deployment. This challenge is further compounded by common characteristics of medical data, including limited sample sizes, severe class imbalance, and feature evolution arising from changes in diagnostic criteria and clinical documentation. To address these issues, we propose Medical Heuristic Learning (MHL), an instantiation of the learning-beyond-gradients paradigm for clinical tabular prediction. Instead of relying on neural network weight updates, MHL uses a large language model (LLM)-driven workflow that integrates statistical probes, medical knowledge probes, rule synthesis, and code-level iterative refinement to optimize a deterministic and executable decision system. The resulting model is expressed not as opaque parameters, but as versioned pure-Python decision rules that are explicitly interpretable, fully auditable, and clinically grounded. MHL also supports continual learning by starting from previously validated rules and iteratively revising them using updated feature information under data drift or feature evolution. Comprehensive experiments on medical datasets show that MHL achieves performance comparable to state-of-the-art methods while maintaining strong behavior in small-sample and highly imbalanced settings. The results further indicate that this explicit rule update mechanism can help alleviate catastrophic forgetting under feature evolution. Overall, these findings suggest that non-gradient-based heuristic systems offer a transparent and adaptable alternative for high-stakes clinical decision support.

Post-traumatic epilepsy (PTE) is a severe complication of traumatic brain injury (TBI), yet early identification remains challenging due to the complex structural and functional alterations it induces in the brain. To address this, we propose a dynamic multimodal Mixture-of-Experts (MoE) framework that integrates functional and structural MRI through time-aware functional-structural encoding and class-conditioned expert routing. Within this framework, modality-specific and cross-modal experts learn complementary representations, while a Modality-Class MoE (MCoE) module dynamically dispatches expert weights according to each classification objective. Experimental results across three binary classification tasks demonstrate that the framework consistently outperforms static fusion baselines, and high-interpretability analyses further reveal meaningful region-of-interest (ROI) interactions. This dynamic multimodal expert framework effectively captures class-dependent brain interaction patterns and provides an interpretable approach for PTE diagnosis and risk stratification.

World-model synthesis aims to turn interaction experience into an internal model of environment dynamics. Existing symbolic approaches often fit observed transitions or mixtures of local rules, but they do not produce a complete executable program that can run independently of the real environment. We present Mind-Studio, a framework that synthesizes executable pygame-style world models from state-action-next-state trajectories using large language models. Mind-Studio combines entropy-selected traces with a lightweight game skill file containing object, action, and static scene information extracted from screenshots. We evaluate synthesis quality with a K-step lookahead fidelity protocol that compares generated world-model rollouts against Real-ALE rollouts from the same state. On Montezuma's Revenge, Mind-Studio improves chosen-action next-state prediction from 0.3% for PoE-World to 48.7% while verifying 5 of 8 subgoals; across Alien, Assault, and Skiing, it achieves stronger branch-level fidelity than prior learned lookahead sources.

Machine interpreting (MI), the live, real-time application of speech translation, has achieved remarkable progress on standard benchmarks, with some systems approaching human parity on textual fidelity. Yet the user experience remains far inferior to interpreter-mediated communication, revealing what we term the accuracy illusion: systems that appear accurate on paper but fail in practice to support smooth, goal-oriented interaction. This paper defines MI as a distinct subfield of speech translation, with its own characteristics and the need for evaluation methods grounded in communicative effectiveness rather than isolated fidelity metrics. Drawing on insights from interpreting studies, we identify critical dimensions of professional interpreting practice that are overlooked by current systems, and consolidate them into three interdependent design priorities for future MI: agency (context-sensitive initiative and repair), grounding (multimodal and discourse-level situational awareness), and experience (adaptive improvement through real interaction). Together, these priorities chart a path toward closing the usability gap and enabling systems that can sustain authentic multilingual communication in real time.

We present GOOSE-M2F, a task-specific adaptation of Mask2Former for the GOOSE 2D Fine-Grained Semantic Segmentation (FGSS) Challenge at ICRA 2026. The GOOSE benchmark spans 64 fine-grained classes across unstructured outdoor terrain with a severely long-tailed distribution, where rare classes occupy fewer than 50 pixels per image. We extend the Swin-Large Mask2Former baseline with three targeted contributions: (1) 200 object queries to eliminate representational saturation; (2) a Feature Refinement Module (FRM) combining ASPP-lite and CBAM dual-attention; and (3) an Auxiliary Supervision Head that delivers direct per-pixel gradients for rare classes. A multi-stage training strategy pairs Distribution-Balanced loss, Rare-Class Copy-Paste augmentation, dynamic IoU-aware re-weighting, and EMA. At inference, a dense sliding-window engine with 2D Gaussian kernel blending and 4-scale TTA adds +10.57%. GOOSE-M2F achieves 70.08% Official Composite mIoU (63.55% fine, 76.61% coarse), placing 3rd on the GOOSE 2D FGSS leaderboard. Code and trained models are publicly available at GitHub: https://github.com/Aditya-Lingam-9000/GOOSE-M2F and Hugging Face: https://huggingface.co/XYZ9843/GOOSE-M2F.

While Large Language Models (LLMs) have substantially advanced text-to-code synthesis, many real programming tasks specify intent through visual artifacts such as screenshots, charts, vector drawings, videos, and interactive states. These tasks require models to connect visual perception to executable programs, because correctness depends not only on syntax but also on layout, data semantics, interaction behavior, and domain-specific constraints that apply after execution. This survey examines Multimodal Code Intelligence, covering systems that generate, edit, refine, or reason with code under visually grounded inputs and outputs. We first formulate the field by the role that code plays in each task, distinguishing code as a rendered artifact, an editable symbolic structure, a scientific representation, an intermediate reasoning trace, or an executable policy or tool interface. We then organize benchmarks and methods into four domains: Graphical User Interface, Scientific Visualization, Structured Graphics, and Frontier Tasks and Frameworks. This taxonomy connects mature artifact-generation problems to emerging agentic and unified settings and allows us to compare how different tasks treat evidence of correctness. Looking ahead, we argue that future research may benefit from four verification-centered directions. Multi-signal validation can combine complementary evidence of correctness, multi-state verification can test behavior across execution trajectories, cross-task transfer testing can probe reusable visual-code skills, and verifiable agent traces can reveal whether agent actions are grounded in visual evidence. Together, these directions may move this field from single-output imitation toward evidence-grounded executable systems. An ongoing project and resources are available on \href{https://github.com/xjywhu/Awesome-Multimodal-LLM-for-Code}{GitHub}.

Where an LLM sits in an agent memory pipeline -- between the recall plane that retrieves stored facts (extensively benchmarked) and the control plane that mutates them via supersede, release, purge (largely untested) -- shapes which forgetting failure modes the system recovers. Comparing thirteen system configurations on a 385-case adversarial surface, we observe three placement regimes with partly complementary coverage: deterministic primitives suffice for lexical/temporal categories but fail canonicalization (5% on identifier-obfuscation, 0% on cross-lingual); inscribe-time LLM recovers canonicalization (100%) but cannot help intent-aware deletion (0% on prefix-collision and compound-fact); a mutation-time hook recovers intent-aware deletion (78-85%) and brightens nearly all categories simultaneously (91.7-93.2% overall, $0.17 per 385-case run, 2.3s/case mutation latency vs. 64-191ms/case deterministic, recall path unchanged). We expose the trade-off via ForgetEval, a 1000-case templated suite plus a 385-case adversarial layer (132 hand-crafted + 253 LLM-drafted oracle-validated) scored by deterministic substring match, paired with a six-method Adapter Protocol with honest N/A scoring that lets heterogeneous memory stores enter in 130 lines. Admission is corroborated by 10-annotator IAA (Fleiss' kappa = 0.958) and a 77-case external-authored subset (four blind contributors) that replicates the canonicalization asymmetry and amplifies the joint-placement lift (+27.8 pt). Production failures are predominantly forgetting failures rather than recall failures, yet existing benchmarks measure only recall. ForgetEval and all adapters are released under MIT.

Kashmiri, an Indo-Aryan language written in a modified Perso-Arabic script, frequently omits diacritic marks in digital text, creating ambiguity and challenging downstream NLP applications. We present Koshur Diacritizer, a ByT5-small byte-level sequence-to-sequence model for restoring diacritics in Kashmiri text. To support this task, we release a publicly available dataset of 23.7k aligned undiacritized diacritized Kashmiri sentence pairs. The proposed framework combines script-aware normalization, alignment validation, and skeleton-preserving inference to ensure reliable restoration while maintaining the original base-letter sequence. Experimental results on a held-out test set achieve a DERm of 0.2012 and a WER of 0.2159. Additionally, evaluation by a native Kashmiri linguistic expert yields a mean accuracy of 77.5%. The dataset, model, and source code are publicly released to provide a reproducible baseline for Kashmiri diacritic restoration and future low-resource language research.

Discharge summaries are crucial clinical documents containing the context of a patient's overall hospital stay, and are routinely reviewed by medical experts for patient readmission, ongoing care, and diagnostic decision-making. When reviewing them, medical experts often must iteratively synthesize information across multiple summaries while verifying the evidence supporting each answer. Although large language models (LLMs) are increasingly explored for clinical question answering, existing benchmarks do not sufficiently reflect this setting: they often evaluate exam-style medical knowledge or focus on single-turn question answering with limited evidence-grounding evaluation. We introduce EHRNote-ChatQA, the first benchmark for evidence-grounded multi-turn clinical question answering over patients' multiple discharge summaries. Built from de-identified MIMIC-IV discharge summaries, EHRNote-ChatQA contains 967 patient-level multi-turn samples spanning one to five notes and 16,072 medical-expert-verified QA pairs (8,036 content questions, each paired with an evidence-grounding question) across eight clinical categories. The benchmark is constructed through an expert-informed pipeline combining discharge-summary structuring schema, expert-curated multi-turn QA templates, and LLM-based generation, followed by review and revision of every single QA sample by 11 medical experts. Benchmarking 22 open- and closed-source LLMs reveals several challenges, including that LLMs struggle more with evidence grounding than content answering, multi-turn errors compound across turns, and single-turn clinical QA performance does not reliably transfer to this setting. These findings establish EHRNote-ChatQA as a rigorous and practical benchmark for evaluating clinical QA systems. The dataset will be made publicly available through PhysioNet credentialed access.

Interpretability is essential for trustworthy medical image diagnosis. However, existing concept-driven interpretable methods have key limitations: Concept Bottleneck Models (CBMs) require scoring all predefined concepts at inference time and for manual intervention, imposing a substantial burden on clinicians, while rationale-based generative approaches often select concepts by class discriminability, which can drift from diagnostic ontologies. To address these issues, we propose Neuro-Symbolic Rule Distillation (NeRD), a framework that produces efficient, ontology-grounded reasoning chains that are sufficient yet non-redundant, without manually crafting diagnostic rules. Experiments on two skin datasets demonstrate strong diagnostic performance and interpretability, and blinded expert evaluation confirms the clinical plausibility of NeRD rationales. Our method further enables a first expert-in-the-loop study for Multimodal Chain-of-Thought-based diagnosis, achieving efficient and effective concept-level intervention.

Collaborative generation, which coordinates multiple diffusion trajectories to extend the capabilities of pretrained priors, has emerged as a powerful paradigm for extending the applicability of diffusion models. Among existing approaches, diffusion synchronization provides a scenario-agnostic solution by introducing general guidance mechanisms. However, current synchronization approaches rely heavily on heuristics and still require task-specific tailoring, which limits their generalizability and performance. In this work, we mathematically derive a synchronization framework based on optimal control, providing a principled explanation of diffusion synchronization. During sampling, we optimize control variables to guide multiple trajectories toward coherent solutions while remaining close to the underlying diffusion prior. Our method operates entirely at test-time without additional training, thereby enabling broad applicability across diverse generation scenarios when combined with strong pretrained priors. We demonstrate consistent improvements over baselines on three representative collaborative generation tasks, covering a wide range of modalities and applications. Beyond performance gains, our work establishes a novel foundation for collaborative generation, opening a principled path toward extending pretrained generative models to new collaborative generation settings.

Organizational knowledge is fragmented across a variety of software systems, tacit expertise, and manual documents that have traditionally been designed for human consumption. As AI systems are increasingly deployed and granted decision-making roles, they require access to this knowledge. This raises two questions: how should organizations store and maintain knowledge so that it remains accessible to both humans and future AI systems, and how should agency be allocated between humans and AI across tasks with different risks and levels of uncertainty? In this position paper, we describe how organizational knowledge evolves and contribute a framework that maps task attributes and knowledge availability to recommended agency allocations and control mechanisms. We illustrate the applicability of the framework on two different manufacturing tasks: a routine operation (visual quality inspection) and a one-off strategic decision (factory location), and conclude with opportunities for future research.

In agentic systems, human-generated data records anchor the value of AI services. Yet cloud compute pipelines centralize processing on remote servers. Data centralization reduces personal data sovereignty and may potentially degrade the quality of service (QoS). Meanwhile, user contributions are diverse in quantity and quality: decentralized records can be biased, noisy, and heterogeneously distributed. To address the data challenge, we study fair token allocation and private data valuation for decentralized and resource-constrained agentic systems. Our approach embeds multi-modal representations in a shared semantic space and releases differentially private (DP) prototypes to preserve utility while reducing semantic leakage. With the DP guarantee, we design a fair token allocation scheme that rewards effective contributions and remains robust to data heterogeneity and AI resource scarcity. Extensive simulations demonstrate improved contribution-based fairness and QoS compared to standard benchmarks. The improved resistance to image reconstruction attacks indicates enhanced privacy for multi-modal personal data.

Adversarial attacks on large language models have limited practical impact despite extensive research. Optimization-based attacks such as Greedy Coordinate Gradient (GCG) (Zou et al., 2023) produce high-perplexity, incoherent suffixes that existing defenses easily detect (Bengio et al., 2024). Moreover, attempting to enforce coherence constraints during optimization often prevents the attack from successfully eliciting the specific targeted response, resulting in low success rates against robust models. Conversely, attacks that maintain coherence often alter the semantic intent of queries; when the model complies with these altered queries, responses fail to address the adversary's original goal. In this work, we introduce Greedy Coordinate Diffusion (GCD), a novel framework that efficiently generates adversarial attacks against safety-aligned models while maintaining low perplexity and high semantic adherence to the adversary's original intent. GCD leverages the generative priors of discrete diffusion language models to guide the search for adversarial suffixes that achieve semantic coherence and adherence. Unlike GCG, GCD does not require direct gradient access, allowing it to operate in a gray-box setting. We show GCD achieves highest ASR while remaining competitive on response-quality scores, and that the constructed adversarial prompts are detected at lower rates than other methods by perplexity-based and guard-model filters.

Open-ended intelligence is the capacity to adapt to novel problems and environments that are substantially different from those in training. A mathematics of open-ended intelligence requires two pillars: first, a minimal set of representational primitives (e.g., states, actions) and algorithmic primitives (e.g., nearest neighbor); and second, an acquired compositional grammar for selection, recursion, and branching that produces sequences of operations and recurring motifs. We formalize open-ended intelligence in terms of the compositional closure induced by a finite primitive set $P$ and a set of composition operators $C$. We characterize properties of the induced closure $\mathcal{L}(P,C)$ that support unbounded compositional generation across families of tasks and worlds. The closure of the two pillars yields infinite adaptive responses across a wide range of settings. The mathematics supports complementary research agendas, including evaluation metrics for explanation and interpretability, and novel architectures where compositional generalization is native. We propose next primitive prediction (NPP) as a novel architectural objective, where training encourages the acquisition of reusable algorithmic primitives and their compositional grammar, such that new solutions are generated through recombination. Given such an objective, curriculum learning and self-play can enable lifelong learning, expanding the closure by discovering reusable primitives and transition motifs across settings. We ground the framework through case studies in physics, evolution, and neuroscience.

Pixel-space diffusion models are trained on full-bandwidth noisy images, yet the useful signal available to the denoiser is strongly frequency dependent. Under rectified-flow diffusion and natural-image power-law spectra, the per-band data-to-noise contour $k^{*}(t) = (1-t)^{-2/α}$ separates a signal-bearing low-frequency region from a noise-dominated high-frequency region at each time $t$. We show that this implicit coarse-to-fine structure is not merely descriptive: it induces a capacity-allocation problem. A standard pixel-space denoiser must discover the moving bandwidth boundary internally and can spend computation on frequency-time regions where the optimal prediction collapses to deterministic baselines rather than data-distribution modeling. To make this boundary explicit, we introduce Spectral Forcing, a parameter-free, time-conditional 2D-DCT low-pass operator applied to the noisy input before the patch embedder. Its cutoff expands monotonically with the diffusion time and becomes the identity at the data endpoint. Through controlled synthetic experiments, we identify the regime in which the operator is beneficial: coarse patch tokenization and data whose high-frequency content is predominantly noise rather than essential signal. On ImageNet-256 with JiT-700M/32, Spectral Forcing consistently improves both FID and Inception Score across different training epochs, demonstrating robust gains throughout training; at finer tokenization, the spectral forcing is still competitive. We further insert the unchanged operator into SenseNova-U1, a unified text-to-image model, where it improves DPG-Bench and GenEval, showing that the input-side spectral prior transfers beyond class-conditional generation. These results suggest a route to capacity-efficient pixel-space diffusion by showing the signal and hiding the noise.

Inverse kinematics (IK) remains a critical bottleneck for real-time robot manipulation. Classical numerical solvers achieve high geometric precision but often suffer from discontinuous branch switching and unstable behavior near kinematic singularities during closed-loop deployment. Meanwhile, learned IK approaches frequently struggle to balance spatial accuracy, motion smoothness, and real-time efficiency, particularly when trained on noisy human teleoperation data. We present \textbf{MimicIK}, a real-time generative inverse kinematics framework that learns smooth and robust joint-space motion priors from teleoperation demonstrations through conditional flow matching. Given the current joint configuration and a target end-effector pose, MimicIK predicts continuous delta-joint commands using an efficient two-step iterative refinement process based on a Minimal Iterative Policy (MIP) backbone. To enforce physical consistency, we further introduce an FK consistency loss, a differentiable forward-kinematics regularization that penalizes task-space deviations from the target pose during training. We evaluate MimicIK on a real-world 6-DOF robot dataset containing 8,848 teleoperation demonstrations. MimicIK achieves a mean position error of 4.65 mm, a 10 mm success rate of 92.01\%, and a trajectory spike rate of only 7.99\%. Compared with a UNet diffusion baseline, our method improves both spatial accuracy and motion smoothness while reducing inference latency from 21.66 ms to 6.74 ms. Furthermore, unlike deterministic MLP baselines that catastrophically diverge under out-of-distribution deployment, MimicIK remains stable near singular configurations and enables robust 20 Hz real-time control on deployment hardware.

Predicting individual panic emotional arousal timing before manifestation is essential for proactive emergency intervention. Existing methods incorporate cognitive elements but none explicitly model the emotional arousal process, making them ill-suited for emotional arousal timing prediction. We argue that grounding prediction in appraisal emotion theory is necessary because it explicitly models this process, but three problems must be solved. (1) Appraisal theory posits that emotion arises from simultaneous evaluation across multiple threat dimensions, yet no prior work fuses these inputs into risk perception. (2) Existing cognitive models lack an Emotion node, decoupling threat appraisal from emotional arousal and forcing emotions to be inferred indirectly from behaviors. (3) Given their generalizable cognitive reasoning, current approaches adopt LLMs as the primary decision-maker, yet overlook the fragility and hallucination-proneness of their outputs. To address these issues, we introduce PanicCognitivePath (PCP), a framework that addresses all three. A Psychological Safety Distance (PSD) model, grounded in psychological distance theory, maps four-domain signals into a unified risk metric as the entry condition for subsequent cognitive reasoning. An explicit Emotion node grounded in appraisal emotion theory is introduced into BDI, forming a Belief-Desire-Emotion-Intention (BDEI) pathway. Agents whose risk metric exceeds the PSD threshold enter this pathway, coupling threat appraisal directly to emotional arousal. The BDEI pathway governs all state transitions while the LLM is confined to parameter estimation for the Belief-to-Desire transition, confining hallucinations to a single step and preventing error propagation. Experiments on Hurricane Sandy show PCP improves arousal timing accuracy by 10.68% over baselines, reduces peak count error to 7.07%.

Sparse autoencoders (SAEs) are standard tools for mechanistic interpretability, but current SAE families are constrained by fixed encoder nonlinearities such as ReLU, JumpReLU, and TopK. This hard-codes a particular sparsity mechanism into the model and can distort the reconstruction-versus-sparsity trade-off. We introduce the Rational Sparse Autoencoder (RSAE), which replaces the fixed encoder activation with a trainable rational function. Rational activations are flexible enough to uniformly approximate the activation primitives used by existing SAE families on compact domains (for TopK, the thresholded gate obtained after a separating top-k threshold is supplied), while also providing a richer function class for adapting to the observed pre-activation geometry. We realise this idea through a two-stage pipeline: an initialisation procedure that copies the pre-trained baseline SAE weights, plugs in rational coefficients obtained by the relaxed Remez exchange on synthetic data, and calibrates the scale parameters along with the rational coefficients; followed by a fine-tuning step under the standard sparsity-regularised reconstruction objective. Empirically, on residual-stream activations of three open-weight language models and across all three baseline activation families, the RSAE strictly improves on it after the fine-tuning step, both on reconstruction-side metrics and on downstream-behaviour metrics, without sacrificing feature-level interpretability under sparse probing. These gains are consistent across host language models, across baseline activation families, and across the full range of baseline sparsity we tested, while the upgrade itself adds only a handful of scalar parameters per autoencoder and runs in minutes on a single consumer GPU.

Multimodal Large Language Models (MLLMs) achieve strong vision-language reasoning, but long visual contexts enlarge the KV cache and increase decoding latency. Existing compression methods rely on observation window attention for stable token-importance estimation, yet this aggregation can dilute sparse visual evidence and discard answer-critical tokens under aggressive compression. Therefore, we identify last-query attention as a complementary source for recovering such evidence, but its answer-irrelevant signals can mislead retention. We propose BACON, a plug-and-play method that calibrates observation window attention with last-query evidence and suppresses isolated noise via intra-layer coherence and inter-layer persistence. Across diverse benchmarks, models, budgets, and compression methods, BACON improves multimodal KV compression by 7.5% on average under the most aggressive budget, with gains up to 30.9%. Our project page is available at https://ryu1ion.github.io/official_BACON/

Knowledge editing systems must update selected facts while preserving nearby but irrelevant behavior. This paper studies this problem in a memory-assisted setting where an edit memory is retrieved at inference time and a parameter-efficient adapter corrects the model's object preference. We argue that the central design question is not only how to write an edit, but also when to suppress it. We introduce \method{}, a route-specialized dual-adapter editor. A relevance router first decides whether a prompt should receive an edit memory. Routed prompts use an edit adapter trained to prefer the new object over the original object; unrouted non-direct prompts use a separate locality adapter trained to preserve or restore the original-object preference. We evaluate \method{} on three 1,000-case protocols, \cf{}, \zsre{}, and \mquake{}, under the same memory protocol and two 7B/8B base models. On Llama-3.1-8B-Instruct, \method{} obtains the best overall probability-preference accuracy on all three benchmarks: 0.8180 on \cf{}, 0.8946 on \zsre{}, and 0.9922 on \mquake{}. The same trend holds on Qwen3-8B. Router ablations show that the relevant memory boundary differs across datasets: a lexical neural router is safest on \cf{}, while BGE embedding routing is better on \zsre{} and \mquake{}. Component and module ablations show that the gain mainly comes from separating edit injection from off-route suppression rather than from simply increasing LoRA capacity.

Robots under autonomous operation may require decisions based on evidence that is no longer visible. We study delayed-evidence tasks, where an early cue disappears before a later decision point, so visually similar observations can require different actions. In these settings, the current observation is not a sufficient state for control. We introduce TRAjectory-routed Causal Evidence (TRACE), a memory framework for visuomotor imitation policies. TRACE stores task-relevant visual and robot-state evidence, such as object identity, target choice, or route-dependent state, in a fixed-size latent memory that remains bounded over long episodes. Instead of indexing memory by raw time or manually provided task labels, TRACE uses path signatures: compact, order-sensitive features of the executed robot-state trajectory. These signatures do not store the visual cue itself; rather, they provide trajectory-conditioned keys for writing and retrieving the evidence stored when the cue was visible. When the robot later reaches an ambiguous observation, the policy conditions on TRACE memory to recover the missing context and choose the correct branch. TRACE attaches through lightweight adapters to policies, without changing the policy backbone, action head, or imitation objective. Across real-world long-horizon manipulation tasks with visually ambiguous branch points, TRACE improves branch selection and task success over alternative baselines, including short-history and recurrent memory. Project page: https://jeong-zju.github.io/trace

End-to-end (E2E) autonomous-driving planners trained by imitation are prone to statistical shortcuts: they associate scene elements that merely co-occur with expert actions (a roadside object, a building facade) with driving decisions, rather than the variables that causally determine them. Such causal confusion silently compromises reliability in long-tail scenarios, and it is difficult to detect, because prevailing open-loop metrics (L2 displacement and collision rate) are dominated by ego status and do not indicate whether a planner depends on spurious cues. Existing remedies based on causal-intervention training require retraining large models and cannot audit a planner that is already deployed. We present CADET, a training-free framework that audits, benchmarks, and repairs spurious reliance in pretrained E2E planners without any parameter update.

Industrial products such as valves and circuit breakers are defined by dense technical specifications that govern procurement, compatibility, and safety across supply chains. These specifications are scattered across multiple heterogeneous product images, including specification tables, nameplates, and technical drawings, yet whether Multimodal Large Language Models (MLLMs) can reliably recover them remains underexplored. To fill this gap, we introduce IndustryBench-MIPU, the first large-scale benchmark for multi-image industrial product understanding, built around structured attribute extraction -- recovering property-value pairs from product images. This task jointly probes text recognition on specification tables and nameplates, visual reasoning over technical drawings, domain knowledge to decode industrial terminology, and cross-image evidence integration to assemble scattered specifications. Concretely, the benchmark comprises 4,559 products across 27,652 images with 103,703 annotations spanning 18 industrial categories, constructed through multi-model consensus and three-tier quality assurance. Evaluating nine MLLMs under both single-image and product-level multi-image settings reveals a stark completeness gap: models achieve high precision (86--94%) but the best recovers only 49.9% of product-level attributes; moving from single-image to multi-image extraction costs 15--34 percentage points of recall. Multi-image completeness, not single-image accuracy, is the core bottleneck. Dataset and code are publicly available.

Large-scale neural network training increasingly relies on matrix-aware optimizers that exploit the structure of weight parameters beyond element-wise adaptation. However, existing matrix-aware methods such as Muon have an underappreciated vulnerability: their core operation, Newton-Schulz iteration, depends critically on input conditioning, yet the raw momentum matrices exhibit severe coordinate-wise scale heterogeneity. In this paper, we first verify this scale heterogeneity through a chi-square uniformity test, showing that intra-matrix scale imbalance is prevalent across Transformer layers and that coordinate whitening effectively corrects it. Motivated by this finding, we propose Zeta, a dual whitening optimizer that applies coordinate whitening and spectral whitening in a strictly ordered pipeline. The ordering is not a tunable choice but follows from a mathematical dependency: coordinate whitening establishes the statistical isotropy that spectral whitening requires to function reliably. We further prove that this dual pipeline strictly reduces orthogonalization error relative to pure spectral methods by improving the condition number of the input. Empirically, Zeta matches or surpasses strong baselines across language modeling (0.6B to 8B parameters), mixture-of-experts architectures, and vision tasks, demonstrating that resolving scale imbalance before orthogonalization leads to faster convergence and better generalization. Code is available at https://github.com/AIGCodeOS/aigcode_zeta_optimizer.

Micro-actions are short-duration, low-amplitude subtle body movements at the whole-body level that can reveal latent intentions, involuntary reactions, and fine-grained affective changes. Our previous MA-52 benchmark has provided an important foundation for micro-action recognition, but it remains limited in scale, scene diversity, task coverage, and evaluation protocols. To advance micro-action analysis toward more realistic and comprehensive settings, we introduce MMA-82, a large-scale multi-domain extension of MA-52. MMA-82 expands the label space from 52 to 82 fine-grained micro-action categories and covers four distinct domains, including laboratory interviews, street interviews, psychiatric patient interviews, and emotion-rich television videos, resulting in 77,856 annotated instances from 454 subjects. Built upon MMA-82, we establish two core tasks: Micro-Action Recognition and Multi-label Micro-Action Detection. For recognition, we further define in-domain and cross-domain protocols, including few-shot and zero-shot settings, to evaluate model robustness, transferability, and generalization. Extensive experiments show that current methods still struggle with realistic micro-action understanding, especially under domain shift, long-tailed category distributions, and complex temporal localization. Beyond benchmarking, we investigate the relationship between micro-actions and emotion, showing that micro-actions are strongly associated with emotional states and provide complementary cues to facial micro-expressions for improved emotion recognition. These results demonstrate that MMA-82 serves as a comprehensive and challenging benchmark for realistic micro-action analysis and a valuable resource for human-centered AI. MMA-82 is available at https://lpynow.github.io/MMA-82-AIM/.

Rapid post-event landslide mapping is essential for disaster response but remains difficult to automate due to extreme class imbalance. This study evaluates whether Clay v1.5, a Geospatial Foundation Model (GFM), can improve pixel-level landslide segmentation on the Landslide4Sense (L4S) benchmark, which contains 3,799 training chips with 14 Sentinel-2 and terrain bands and approximately 2% positive pixels. We compare three strategies: Clay as the primary encoder with multi-scale residual terrain fusion, a U-Net backbone augmented with Clay semantic context at the bottleneck, and a standard U-Net baseline. The hybrid U-Net + Clay model with two-stage Low-Rank Adaptation (LoRA) achieved the best test F1 of 64.5 +/- 1.8% over three seeds, surpassing the Clay-only backbone (55.2 +/- 3.6%) and the U-Net baseline (59.9%). Clay as a standalone encoder underperformed the U-Net due to the absence of multi-scale skip connections, but its pretrained representations consistently improved performance when injected as auxiliary context. These findings suggest that GFMs are most effective for landslide detection when they complement spatially detailed convolutional architectures rather than replace them.