Jingsong Liu, Youssef Wally Han Li, Weiwei Zhou, Jing Dai, Elisabeth Wetzer, Peter J. Schüffler
The growing adoption of digital pathology has rapidly expanded collections of hematoxylin-and-eosin whole-slide images, yet efficient navigation through these archives remains challenging, as metadata is often inconsistent or unavailable. Content-based image retrieval addresses this challenge by identifying similar cases directly from visual representations instead of manual annotations or metadata. Existing retrieval systems have demonstrated promising performance but largely rely on handcrafted indexing pipelines and predefined sampling strategies, constraining scalability and generalization. Here, we present CLEAR-WSI (Constant-Length Embedding for Automatic Retrieval of Whole-Slide Images) as a scalable, diagnostic-driven slide-retrieval framework, validated on both public and clinical data. Using a fixed-length slide embedding, CLEAR-WSI refines retrieved candidates with either a predicted diagnostic label or the ground-truth label, depending on the retrieval setting. Newly processed cases are subsequently added to the expanding memory bank, allowing the retrieval archive to grow over time. On a comprehensive benchmark consisting of two public and one private clinical datasets representing four retrieval tasks, CLEAR-WSI improves Precision@5 over the strongest evaluated baseline, with gains on CAMELYON16 (3.10), BRACS-1 (2.75), BRACS-2 (1.95), and X-HER2 (3.75). Additional ablation and latency studies further examine the contribution of each component and highlight its potential for scalable deployment.
Multimodal large language models (MLLMs) remain prone to visual hallucinations when correct responses depend on subtle visual details. One line of post-training methods constructs visual contrasts through image retrieval, generation, or editing to improve fine-grained perception. However, these contrasts are not built around the model's own hallucinations, making it difficult to ensure that supervision targets its actual perceptual confusions. Another line suppresses hallucinated outputs through response correction or preference learning under fixed image conditions, but leaves the design of supervision signals on the image-input side underexplored. Inspired by how humans correct misconceptions, we introduce DreamRealm, a post-training framework that explores diverse latent hallucinated responses, minimally edits original images to make them valid, and trains the model to assign each response to its supporting image. Our data construction pipeline, RealmForgePipe, mines observed and latent hallucinations from a target model and materializes them through image editing. Using hallucinations mined from Qwen3.5-9B, we construct the Realized Hallucination Dataset (RealmSet). Our training objective, RealmMatch, constructs a likelihood score matrix across image--response combinations and combines supervised learning with bidirectional contrastive losses to reinforce correct matches and suppress mismatches. Compared with the original Qwen3.5-9B model, DreamRealm improves performance across six visual perception and hallucination benchmarks, including gains of 7.26 percentage points in HallusionBench accuracy and 11.34 percentage points in MMVP pair accuracy. Together with ablation studies on data construction and training methods, these results demonstrate that hallucination assignment is a promising paradigm for improving visual perception and mitigating hallucinations.
Protein structure prediction has largely focused on native conformations, leaving the continuum of unfolded states, folding intermediates, and native structures difficult to access. We present FoldScape, a sequence-conditioned deep learning framework for generating protein conformational ensembles at specified folding fractions $Q$. Trained on approximately one million conformations from 10,000 protein sequences, FoldScape combines coarse-grained folding and all-atom mechanical-unfolding trajectories. It reconstructs three-dimensional $C_α$ chains through hierarchical sliding-window regressions of inter-residue distances, chain-level assembly, and a rectified-flow decoder. By conditioning on folding progress, FoldScape generates conformations from unfolded states through intermediates to native-state structures without protein-specific molecular dynamics at inference. In matched-sample comparisons, FoldScape achieves median conformational-space coverage of 0.989, 0.722, and 0.877 along radius of gyration, fraction of native contacts $Q$, and secondary-structure fraction, respectively, versus 0.05, 0.05, and 0.10 for seed-varied AlphaFold 3 ensembles. Coverage advantages are consistent across the evaluated proteins and robust to binning choices. Reconstruction accuracy remains approximately stable across the evaluated stages of a folding trajectory, and stride-one crop-and-superpose inference extends predictions to chains several times longer than the single-pass limit. Analyses of mechanical-unfolding trajectories show residue-resolved pathways, reproducible extension-conditioned observables across pulling velocities, and geometrically coherent segments in partially unfolded chains. FoldScape extends sequence-based structure prediction beyond a single folded endpoint to controllable conformational-ensemble generation across protein folding.
We study the random-cluster model on locally treelike graphs at low temperatures. This is a model of dependent percolation parametrized by an edge probability $p \in (0,1)$ and a clustering weight $q \in [1,\infty)$, generalizing independent Bernoulli percolation ($q=1$) and closely related to the classical ferromagnetic Ising and Potts spin systems at integer $q$. For $q>2$, at parameter $p$ below the tree uniqueness threshold $p_{\mathsf u}(q,d)$, it is expected that sampling is easy and local Markov chains mix rapidly on all graphs with degree bounded by $d$. On typical graphs, the same is predicted at $p>p_{\mathsf s}(q,d)$, where $p_{\mathsf s}(q,d)$ is a second uniqueness transition point on the $d$-regular wired tree. Our result confirms this prediction for random regular graphs, showing that the random-cluster Glauber dynamics mix in time $O(n^{1+o(1)})$ on a random $d$-regular graph for all fixed $q>2$ and $p>p_{\mathsf s}$. For comparison, the best prior sampling results for these models that held for all $p>p_{\mathsf s}$ required $q\gtrsim \log d$ \cite{BGPZ26}. The key difficulty is that simply utilizing the contraction of cavity messages on $d$-regular trees does not suffice to cover the entire regime $q>2$. To address this issue, motivated by \cite{LS26}, we develop a careful bootstrap argument to show the shrinkage of disagreement under grand coupling. The core ideas of the proof were generated using GPT-6 and the Pharos system; the proof was then refined and simplified by the author.
Video multimodal large language models achieve strong results on existing benchmarks, but answer accuracy alone does not establish whether they can locate the evidence needed to answer a question. We introduce VideoZeroBench, a challenging long-video benchmark with manually annotated question-answer pairs spanning 13 video domains. Questions target fine-grained cues, fleeting events, and evidence distributed across multiple segments. Temporal intervals and key-frame boxes are annotated where applicable. All questions undergo two rounds of cross-verification for answer validity and evidence quality. Our five-level diagnostic protocol compares answering with and without evidence hints, then combines answer correctness with independently evaluated temporal and spatial grounding. Across 19 evaluated models, the best standard QA accuracy is 24.8% (Level-3), achieved by Gemini-3.7-Flash. No model exceeds 1.8% when correct answers and accurate spatio-temporal localization are jointly required (Level-5). Analyses of atomic abilities, evidence spans, input modalities, and thinking-with-videos inference further characterize where the evaluated systems struggle. These findings motivate more precise evidence search and localization for long-video question answering. Our code and data are publicly released.
Muon has recently shown strong empirical performance in training Transformer-based large language models (LLMs), yet its optimization dynamics remain poorly understood from a theoretical perspective. In this paper, we analyze the training dynamics of Muon for next-token prediction (NTP) in Transformers. The training procedure consists of two phases: the preprocessing phase learns the output-value component, while the main phase learns the key-query component. For the preprocessing phase, we characterize the optimization trajectory exactly and show that the normalized spectral margin is optimal after each update. For the main phase, we establish nonasymptotic convergence rates for the training loss and the normalized margin gap under softmax attention. These results suggest that Muon's update rule can help models efficiently learn which tokens matter for prediction, offering a possible explanation for its strong performance in large language model training. The learned weight structure also yields a generalization guarantee for unseen contexts. Furthermore, we validate our theoretical insights with experiments on next-token prediction tasks.
Autoregressive video diffusion models (AR-VDMs) enable real-time, long-horizon video generation but degrade rapidly once the inference horizon outruns the training horizon, due to compounding exposure bias. To stabilize long rollouts, recent methods heuristically adopt frame sink, keeping initial frames as a persistent model context, inspired by attention sinks in LLMs. We revisit this design and find that the mechanism by which initial frames stabilize long-range consistency in AR-VDMs is not an LLM-style attention sink; instead, long-range consistency is anchored through a sparse subset of attention heads, while the majority of heads predominantly capture local temporal dynamics. This observation reveals an over-conditioning pathology in existing frame-sink strategies: by universally exposing every head to the pinned initial-frame context, the model becomes prone to collapsing motion toward the low-dynamic modes of the teacher during model distillation. We propose Disen-Forcing, a distillation-time plug-in for AR-VDMs that disentangles and reinforces this head specialization through disentangled context routing and motion-disentangling regularization. On long-horizon video generation benchmarks, Disen-Forcing simultaneously improves long-range consistency and motion dynamics of base AR-VDMs, a trade-off that sink-based methods fail to balance.
We present Youtu-Parsing-Omni, an omni-modal parser that converts documents, images, audio, and video into structured records grounded in space and time, providing a perception layer for agents that read, locate, and cite evidence across modalities. Its design rests on one encoder for every modality and one schema for every task. Youtu-Omni-Encoder, initialized from a pre-trained language model to strengthen text reading, encodes all modalities with shared weights in a unified space-time coordinate system, and a few fusion layers let video frames attend to concurrent audio, bringing cross-modal fusion and temporal alignment into perception itself. OmniSchema unifies the outputs of all ten parsing tasks, allowing a single self-evolving post-training procedure, OmniSchema-Aware On-Policy Distillation (OSAD), to provide a shared token-role-based objective across tasks. In each round, the model is trained on its own rollouts against token-level targets from a reference-conditioned copy of itself, which is then refreshed with the trained weights, so that the parser and its supervision co-evolve. With a single 5 B-parameter checkpoint, Youtu-Parsing-Omni achieves state-of-the-art performance on OmniDocBench v1.6 (96.96 Overall) and leads all open-weight models on OmniParsingBench (75.08 Avg.), with its advantage concentrated on audio and video. It also achieves the best overall chart-parsing score (95.02) and the highest exact chemical-fingerprint agreement among compared systems, and matches a dedicated optical-music model in character-level transcription of typeset sheet music while staying far ahead of general-purpose systems. Code and model weights are publicly available.
Proximity gaps and correlated agreement underpin the soundness of interactive oracle proofs. We establish these properties for general one-point algebraic geometry (AG) codes, extending corresponding results for Reed-Solomon codes. For codes of length $n$ over $\mathbb F_q$ with designed distance $n-m$, write $ρ=m/n$ and fix $ρ\in(0,1)$. For polynomial curves of degree at most $\ell$ and any $0<\varepsilon<1$, we prove correlated agreement with multiplicative slack $\varepsilon$ and soundness $O(\ell(\ell+1)/(\varepsilon q))$ whenever $e\le(1-\varepsilon)(n-m)/2$, with no additional genus penalty. This gives asymptotically good AG code families from Garcia--Stichtenoth tower of unbounded length over a fixed field with soundness bounds independent of $n$, beyond the RS restriction $n\le q$. We also establish mutual correlated agreement up to radius $(1-\sqrtρ-η)n$, with soundness $O(\ell n/(η^3q))$ for $0<η<1-\sqrtρ$. Our half designed distance proof combines auxiliary divisors, simultaneous Berlekamp-Welch interpolation, and error-incidence counting to obtain a common polynomial representation without a Polishchuk-Spielman divisibility lemma. Finally, two counterexamples on the Garcia-Stichtenoth tower show limitations beyond half the minimum distance, at radii below the Johnson bound and with positive slack. The first requires a positive constant soundness lower bound independent of $q$ and $n$, precluding a uniform $O(1/q)$ guarantee throughout the Johnson regime. The second fixes the field and lets the relative gap above half distance tend to zero while ruling out correlated agreement for every soundness parameter $σ<1$, with positive limiting rate and relative error radius.
Atmospheric chemical mechanisms are commonly distributed in the language of the Kinetic PreProcessor (KPP), which generates efficient Fortran or C code but does not fit into the differentiable programming frameworks in which machine-learning models are developed. We present KPPax, a kinetic preprocessor that generates differentiable, just-in-time compiled, and vectorisable chemistry code in JAX from mechanisms written in the KPP language. KPPax evaluates the rate law with sparse operations in log space, which keeps it differentiable, and integrates the stiff system with the implicit solvers and the forward- and reverse-mode adjoints of the \texttt{diffrax} library, so that exact sensitivities are obtained without hand-derived adjoint code. We demonstrate two applications with the POLLU benchmark mechanism. In the first, the Fisher information matrix computed from an automatically differentiated Jacobian predicts, before any fit, which of the 25 rate coefficients a flow-tube-like design can identify, and noisy synthetic estimation confirms the predicted uncertainties of the nine identifiable rate coefficients to within about 30\% while revealing clusters of fast reactions of which only ratios are constrained. In the second, a neural-network correction of an emission inventory is trained offline on per-scenario 4D-Var retrievals and then fine-tuned online through the chemical model, which reduces the error of the biogenic correction by a factor of 2.4--5 and recovers information on low-temperature emissions that the individual retrievals lose. KPPax makes KPP mechanisms directly usable for parameter estimation and hybrid physics--machine-learning modelling.
Ziyu Cao, William Talbot, Song Li, Pakpong Chirarattananon, Fredrik Gustafsson, Marco Hutter, Kailai Li
We present RESPLE2, a recursive multi-sensor state estimation framework using non-uniform B-splines for 6-DoF motion representation. By exploiting the knot-clamping boundary property, we propose a theoretically grounded mechanism for exact control-point prediction directly informed by inertial kinematics along the trajectory, with knot placement adapted to motion and measurement availability. Within an iterated extended Kalman filter, we perform this prediction using a subset of inertial measurements. A subsequent reset step preserves trajectory consistency as the knots extend. The update step further incorporates heterogeneous sensor measurements at their native timestamps through continuous-time observation models, supporting flexible combinations of LiDARs, cameras, inertial measurement units, and leg kinematics. We further develop a lightweight global optimization module that directly refines non-uniform B-spline control points by incorporating loop-closure constraints derived from LiDAR-visual cues. We conduct extensive evaluations on public and experimental datasets across handheld, tracked, legged, and hybrid aerial-legged platforms, covering highly dynamic maneuvers with extreme accelerations. Compared with state-of-the-art multi-sensor odometry systems, RESPLE2 achieves competitive or superior estimation accuracy and robustness in the corresponding sensor configurations, particularly under aggressive motion, while maintaining real-time operation. We publicly release our code and experimental datasets at https://github.com/ASIG-X/RESPLE2.
Kailai Feng, Haoxiang Rao, Yuxiang Wei, Bo Chen, Junsong Li, Yang Pan, Chengkun Cai, Jagmohan Chauhan
Developing on-device image models requires considering practical user needs, task types and device resource constraints. However, existing benchmarks provide limited guidance on identifying capability gaps between on-device models and server-side models. We introduce MoI-Bench, a benchmark grounded in mobile user needs, comprising 64 subtasks and 1150 samples. It defines 16 operations, such as removal and replacement, and eight visual target categories covering content such as portraits and text. These form two orthogonal axes whose applicable combinations define specific tasks, such as removing facial makeup. To better guide model improvement, we further propose MoI-Eval, which introduces task-specific criteria covering instruction following, visual quality and detail preservation. We further construct sample-level evaluation questions and aggregate their scores into overall quality and capability profiles along the operation, target and criteria axes. We evaluate representative on-device and server-side models and measure mobile deployment costs. Results show that this decoupled Operation x Target design provides finer-grained guidance for improving compact models. For example, it highlights the need to prioritize specific capability gaps (e.g., text-replacement) over broad application-level goals (e.g., improving text editing). Unlike conventional benchmarks that rank models using a single score, we introduce a two-dimensional Quality-Deployability capability map to assess models for practical mobile use. The map shows that current on-device models can achieve efficient generation or editing but expanding their task coverage deserves greater attention than further improving deployment efficiency. Further progress requires improving editing quality, particularly identity preservation and text rendering, while broadening task coverage.
Zhi-Song Liu, Wenqing Peng, Ammar Kheder, Risto Makkonen, Michael Boy
Earth-system models simplify atmospheric chemistry into condensed mechanisms, because near-explicit mechanisms such as the Master Chemical Mechanism (MCM) are too costly to solve in every grid cell. Neural emulators could lift this limit, but most advance the chemistry in small time steps and cover only a few species or a reduced mechanism. Here we introduce FNO-MoE, a Fourier neural operator with a mixture of experts that predicts the full near-explicit chemical state over 30 minutes in a single call. The same network also guides its own training, using the disagreement among its experts to select the most informative box-model simulations. At equal cost, this active sampling yields a more accurate emulator than random or evenly spread sampling. Fine-tuning on longer trajectories keeps it stable over multi-hour runs in which models trained on short simulations diverge. The resulting emulator reproduces the evolution of thousands of species with high fidelity, remains accurate for unseen chemical compositions and runs orders of magnitude faster than the numerical solver. These results open a new direction for emulating complex atmospheric systems: neural operators that predict the full chemical state over long windows and choose their own training data. Training under changing meteorology is the next step towards bringing near-explicit chemistry into Earth-system models.
Zhi-Song Liu, Helmi Toropainen, Ammar Kheder, Michael Boy
Decisions about heat, frost and flooding are made locally, yet reanalyses and climate models resolve tens of kilometres. Generative downscalers add kilometre-scale detail with uncertainty, but are tied to a grid, costly to sample and lack a tractable likelihood. Here we introduce CD-FFM, a functional flow-matching model that treats the high-resolution field as a continuous random field that one model can query at any location and resolution. Its velocity acts pointwise, so spatial variability comes from a noise prior fitted to the observed spectrum and every observation has an exact likelihood. Downscaling ERA5 to 1 km daily temperature over Fennoscandia, CD-FFM lowers the continuous ranked probability score by 27-31% against regression and 12-18% against a forty-times larger neural operator in each of three test years, and outperforms a CorrDiff-style diffusion model while sampling about fifty times faster. The same model transfers to precipitation and makes large kilometre-scale ensembles affordable.
Hesong Li, Ying Fu, Ziqi Wu, Ruiwen Shao, Tao Zhang
Scanning Transmission Electron Microscopy (STEM) enables the observation of atomic arrangements at sub-angstrom resolution and plays a key role in the discovery and characterization of advanced materials. Raw STEM images are often noisy and blurred due to multiple degradations. Although recent deep learning methods improve STEM images, their performance remains limited under severe degradations. In this work, we propose a spatial-frequency learning framework for STEM image enhancement. Specifically, we design a spatial-frequency fusion network that exploits periodic atomic arrangements in the frequency domain and restores local atomic details in the spatial domain. Spatial- and frequency-domain features are enhanced separately and then fused. We also present a STEM noise calibration method for synthesizing more realistic STEM images. The parameters of scan noise, pointwise noise, and background noise are estimated through statistical analysis and fitting of real STEM images. Using these calibrated parameters, we construct a diverse dataset containing a wider range of atomic arrangements, imaging modes, and degradations. Random masks combine different atomic arrangements to simulate interfaces between multiple materials. Experimental results demonstrate that our method achieves state-of-the-art enhancement performance on synthetic datasets and real images of advanced materials while improving atomic localization accuracy.
Yingbo Hao, Song Liu, Guanghua Yu, Zhaojing Zhou, Jianchen Zhu, Yi Zou
Multi-turn agent reinforcement learning (RL) relies on long closed-loop rollouts, making generation and KV-cache costs a major bottleneck. Quantized rollout reduces these costs, but its low-precision executor also serves as the behavior policy and can induce execution mismatch with the learner. Existing quantized-RL methods are designed primarily for single-turn generation. They align the learner forward with quantized rollout or correct the current-token rollout-learner ratio, without modeling the policy history that produces each agent state. Even after quantization alignment, residual rollout-learner mismatch accumulates along the state-reaching policy history, so a current action may look locally aligned despite a mismatched history. We introduce InterQ-RL (Interaction-aware Quantized Reinforcement Learning) to correct this post-alignment mismatch. It applies bounded execution-prefix correction over the state-reaching policy history, allocates correction strength across action segments by effective support under a fixed displacement budget, and requires no additional model forward pass. We evaluate dense and mixture-of-experts (MoE) backbones from 4B to 30B on ALFWorld, WebShop, and ScienceWorld, with all quantized arms sharing the same rollout with 8-bit weights, activations, and KV cache (W8A8-KV8). Across four environment-backbone settings, InterQ-RL achieves the strongest quantized endpoint result, matching BF16 to within 0.7 points and exceeding it in two settings. It matches or exceeds the learning speed of alignment alone while avoiding its late performance regression.
Multi-modal video fusion aims to generate temporally coherent videos with salient targets and rich structural details by integrating complementary information from heterogeneous sensors. However, existing video fusion methods commonly rely on optical flow estimation and feature warping for inter-frame alignment and cross-frame information aggregation to maintain temporal consistency, thereby introducing considerable computational overhead and making the fusion performance vulnerable to flow estimation errors. To address these limitations, this paper proposes an efficient multi-modal video fusion network with dynamic region-guided temporal interaction. Specifically, the proposed spatio-temporal interaction module leverages the Test-Time Training paradigm to model temporal dependencies with linear complexity. Building on this design, the module restricts temporal modeling to key dynamic regions in the video sequence, as static background regions provide limited temporal cues. By excluding low-variation tokens from unnecessary temporal interaction, this selective strategy contributes to improved fusion quality and temporal consistency. In addition, leveraging the strong ability of pretrained MAE features to capture global structural dependencies, we introduce a feature-level temporal consistency constraint to suppress local flickering by aligning the inter-frame variation directions of the fused and source videos. We further propose a new MAE-based temporal metric that quantifies directional distance between their temporal variations within dynamic regions. Experimental results demonstrate that, compared with existing state-of-the-art methods, our method achieves superior fusion quality and temporal consistency on multiple benchmark datasets, while exhibiting high computational efficiency under edge-device deployment settings.
Jiayu Tang, Guowei Peng, Qiuhao Xie, Yuning Yang, Xiurui Xie University of Electronic Science and Technology of China, Guisong Liu
The “Fine-Tuning-as-a-Service” paradigm exposes large language models to catastrophic safety degradation from less harmful samples. Alignment-stage defenses address this by proactively injecting adversarial perturbations to bolster the model’s inherent robustness against harmful drift. However, existing methods rely on perturbation directions that often conflict with harmful gradients, inadvertently facilitating the acquisition of malicious features rather than suppressing them. To address this issue, we propose Orthogonal and Adaptive Safety Alignment Strategy (OASIS) to mathematically decouple safety enforcement from harmful feature acquisition. By projecting perturbations orthogonal to harmful gradients and concentrating optimization on adaptively selected safety-critical layers, OASIS effectively resolves directional conflicts while maximizing parameter efficiency. Extensive experiments on four LLMs across three datasets (SST2, GSM8K, and AGNews) demonstrate that OASIS reduces the Harmful Score by approximately 60% compared to competitive baselines, while maintaining stable downstream task utility.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表