Sign in

Category: Other

  • SpecFit: Not All Draft Tokens Cost the Same in Offloaded MoE Inference

    SpecFit: Not All Draft Tokens Cost the Same in Offloaded MoE Inference

    arXiv:2610.15417v1 Announce Type: new Abstract: Mixture-of-experts (MoE) models activate only a few experts per token, but when their expert weights are offloaded to host memory, host-to-device (H2D) expert transfers enter the critical path of decoding. Speculative decoding verifies several draft tokens in one forward pass, yet in an MoE model each verified token activates its own experts, so a larger draft tree transfers more experts. Because of this H2D cost, speculative decoding commits several tokens per verification cycle yet is sometimes even slower than autoregressive decoding. Based on this observation, we propose SpecFit, which chooses the verified draft subtree to maximize predicted throughput rather than acceptance length. SpecFit estimates the expected number of committed tokens from draft path probabilities and predicts cycle time from the node count and the per-layer union of the experts the subtree activates; this union is a monotone submodular coverage function, so each node pays only for the experts it adds. An exact search over ancestor-closed subtrees returns the subtree with the highest predicted throughput; run after each draft level, the same search also decides when to stop drafting, and the unmodified target verifies the result. On CNN/DailyMail summarization and a GSM8K subset with Qwen3-30B-A3B, SpecFit decodes 21.6% and 7.0% faster than EAGLE-3 with the same expert cache and 1.97x and 1.72x as fast as vanilla EAGLE-3. Code is available at https://github.com/Chlience/SpecFit.

    Fallback brief based on the available RSS feed; no unsupported details have been added.

    Source: arXiv
    Original article: https://arxiv.org/abs/2610.15417

  • Hidden in Plain Sight: Advancing Harmful Content Understanding via Perception-Aware Adversarial Self-Play

    Hidden in Plain Sight: Advancing Harmful Content Understanding via Perception-Aware Adversarial Self-Play

    arXiv:2610.15449v2 Announce Type: new Abstract: Despite significant advances in language understanding, Large Language Models (LLMs) can still miss meaning hidden in plain sight, failing to identify harmful intent in expressions readily understood by humans. Malicious users exploit this gap by strategically disguising harmful content to evade automated moderation, posing a serious threat to online platform safety. The vulnerability reflects a fundamental cognitive asymmetry between human readers, who integrate contextual, phonetic, and visual cues to decipher the intent behind deliberate obfuscation, and text-based LLMs, which rely primarily on token sequences and have limited exposure to evolving evasions. To bridge this asymmetry, we propose PASP (Perception-Aware Adversarial Self-Play), which combines multi-perceptual views with adversarial self-play to improve understanding of harmful intent as evasions evolve. Specifically, PASP constructs textual, phonetic, and visual views of each input, preserving its original context while exposing pronunciation and glyph cues obscured in token space. During adversarial self-play, an attacker uses defender feedback to generate intent-preserving evasions, while the defender is trained using an intent-invariance reward to understand harmful intent across changing surface forms. Extensive experiments show that PASP reduces adaptive attack success rates by up to 73.71% and 74.49% on ChineseHarm-Bench and EvoHarmBench, respectively, preserves model capabilities with minimal degradation, and effectively interprets real-world evasive expressions in e-commerce.

    Fallback brief based on the available RSS feed; no unsupported details have been added.

    Source: arXiv
    Original article: https://arxiv.org/abs/2610.15449

  • Fewer Bits, More Tries: Recovering Capability Under Memory Constraints

    Fewer Bits, More Tries: Recovering Capability Under Memory Constraints

    arXiv:2610.15372v1 Announce Type: new Abstract: Large language models are generally more capable than smaller ones, but their memory requirements can make local deployment difficult. We ask whether part of this capability gap can instead be recovered by running a smaller or more compressed model multiple times and selecting among its outputs using public or model-generated tests. Single-user local inference can leave hardware underutilized, allowing parallel candidate generation to trade additional test-time computation for accuracy at modest latency cost. Across four code-generation benchmarks, verified retry recovers substantial accuracy lost to model size or compression. A 7B Qwen model with eight parallel attempts essentially matches a 32B model on MBPP+ and exceeds it on the other three benchmarks, while using about $4\times$ less model memory, running up to $6\times$ faster, and using less energy, as measured on Nvidia GPU. We observe similar tradeoff across model sizes, compression levels and hardware choices, although exact benefit depends on the hardware profile and problem setting. These results show that model capacity can be traded for verified test-time computation when correct solutions remain reachable within a modest retry budget.

    Fallback brief based on the available RSS feed; no unsupported details have been added.

    Source: arXiv
    Original article: https://arxiv.org/abs/2610.15372

  • Analyzing the Dynamics of WSBM-based Echo State Networks using InputDSA

    Analyzing the Dynamics of WSBM-based Echo State Networks using InputDSA

    arXiv:2610.15380v1 Announce Type: new Abstract: Bridging the structural gap between biological brains and artificial neural networks presents a pathway to altered computational dynamics. This paper introduces a Reservoir Computing framework that replaces random networks with biologically realistic modularity by applying the Weighted Stochastic Block Model (WSBM) to Echo State Networks (ESNs). This topological parameterization constructs reservoirs reflecting latent functional sub-circuits. Furthermore, the integration of Input Dynamical Similarity Analysis (InputDSA) provides a mechanism to isolate and compare how structural priors modulate intrinsic recurrent dynamics versus external inputs. By evaluating these architectures, we establish that structural diversity does not necessarily predicts behavioral diversity across the selected network subset. Furthermore, our demixed analysis confirms Anna Karenina principle of dynamics, demonstrating that environmental constraints enforce dynamical convergence in successful agents, while failing networks degenerate into idiosyncratic trajectories.

    Fallback brief based on the available RSS feed; no unsupported details have been added.

    Source: arXiv
    Original article: https://arxiv.org/abs/2610.15380

  • A Decade of PhotoMappers: A Longitudinal Study of Crowdsourced Disaster Photos Geolocalization with Vision-Language Models and Geospatial Reasoning

    A Decade of PhotoMappers: A Longitudinal Study of Crowdsourced Disaster Photos Geolocalization with Vision-Language Models and Geospatial Reasoning

    arXiv:2610.15403v1 Announce Type: new Abstract: Crowdsourced disaster imagery provides timely, ground-level observations when conventional disaster mapping is delayed or incomplete, yet its geographic value depends on where observations are produced, how they are interpreted, and whether they can be reliably located as actionable Geospatial Intelligence (GEOINT). Using more than a decade of PhotoMappers, a crowdsourced disaster-mapping initiative, this study examines the changing geography of disaster information and asks how emerging GeoAI models compare with human geographic interpretation. We combine longitudinal spatial analysis with cross-view geolocalization using street view imagery (SVI) and remote sensing imagery (RSI), and evaluate vision-language models (VLMs) reasoning through 30 indicators spanning geographic cues and reasoning evidence. The results reveal a geographically and temporally uneven archive shaped by changing disaster types, Community Lifelines, and operational priorities. Integrating VGI, SVI, and RSI with Sinkhorn re-ranking achieves 40.90% R@1, 61.10% R@5, and 69.45% R@10, while yearly R@1 varies markedly from 25.00% to 65.07%. Geographic reasoning is similarly uneven: Qwen3.6-27B achieves an overall agreement of 67% compared with 51% for InternVL3-8B, with both models showing greater difficulty with localized and place-specific evidence than broad spatial configurations. We argue that the value of GeoAI lies less in replacing volunteered geographic interpretation than in complementing it, scaling candidate retrieval while retaining human judgment for ambiguous and consequential cases. This provides a pathway toward more human-centered and responsible GeoAI for disaster mapping. All data and code are openly available at https://github.com/bobleegogogo/PhotoMappers-VLM.

    Fallback brief based on the available RSS feed; no unsupported details have been added.

    Source: arXiv
    Original article: https://arxiv.org/abs/2610.15403

  • SuperAI: Merging Capabilities Across Heterogeneous Foundation Models

    SuperAI: Merging Capabilities Across Heterogeneous Foundation Models

    arXiv:2610.15322v2 Announce Type: new Abstract: Enabling the reuse and composition of knowledge and capabilities distributed across foundation models is an important step toward building scalable artificial intelligence (AI) systems. However, architectural differences and incompatible representations make it difficult to transfer and integrate these capabilities through a unified cross-model interface. To address this challenge, we propose SuperAI, a cross-architecture multi-model fusion framework that integrates complementary capabilities from multiple source models into a host model through orthogonal alignment of low-dimensional task-relevant representations. Specifically, we first extract task-relevant representations from the source models and the host model using shared anchor samples, and apply principal component analysis (PCA) to construct low-dimensional functional subspaces with the same dimensionality. Next, we use orthogonal Procrustes mappings to align the source and host coordinate systems, establishing cross-model correspondences while preserving the internal geometry of each source subspace. Finally, we train a lightweight residual multilayer perceptron (MLP) with joint supervision on task outputs and source-model increments, learning input-dependent residual increments that are injected into the host model for multi-model capability fusion. Experiments on three host models across 15 capability domains show that SuperAI achieves higher average domain scores than the evaluated cross-architecture fusion methods and better preserves general capabilities as the number of source models increases. Ablation studies further support the effectiveness of orthogonal alignment, nonlinear residual mapping, and joint supervision.

    Fallback brief based on the available RSS feed; no unsupported details have been added.

    Source: arXiv
    Original article: https://arxiv.org/abs/2610.15322

  • MultiFlow: Vision-Driven Multimodal fMRI Encoding via Stimulus-Informed Source Flow Matching

    MultiFlow: Vision-Driven Multimodal fMRI Encoding via Stimulus-Informed Source Flow Matching

    arXiv:2610.15339v1 Announce Type: new Abstract: Current fMRI encoding models frame prediction as deterministic regression, yet identical stimuli evoke a distribution of neural responses (arising from physiological noise, hemodynamic variability, and cognitive fluctuations) that a single point estimate cannot capture. This work introduces MultiFlow, a generative fMRI encoding framework based on Conditional Flow Matching (CFM) and a formulation termed Stimulus-Informed Source Flow Matching (SISFM). SISFM generalises CFM by conditioning the source distribution on the stimulus: a learnable temporal convolution shapes the source mean with an HRF-length receptive field, while a variance head shrinks its scale, shortening the expected transport path. SISFM is instantiated with a Diffusion Transformer (DiT-X) decoder parameterising the velocity field, with sub-layer Attention Residuals, heterogeneous fusion of ten frozen vision, audio, language, and audiovisual backbones, and per-subject Yeo-network heads. On the Algonauts 2025 dataset, a single MultiFlow checkpoint recovers most of the benefit of large deterministic ensembles: it trails the 1000-model TRIBE ensemble in-distribution by 0.003 PCC (0.3163 vs. 0.3195), and the strongest in- and out-of-distribution ensembles (VIBE-20, MedARC-50) by 0.010 and 0.009 PCC, using a single trained model rather than an ensemble of 20-1000. The same checkpoint also provides stimulus-conditioned uncertainty estimates and sample-based diversity.

    Fallback brief based on the available RSS feed; no unsupported details have been added.

    Source: arXiv
    Original article: https://arxiv.org/abs/2610.15339

  • TRACE: Temporal Risk Assessment and Concept Editing for Safer Language Generation

    TRACE: Temporal Risk Assessment and Concept Editing for Safer Language Generation

    arXiv:2610.15365v1 Announce Type: new Abstract: Large language models (LLMs) can produce harmful content while generating a response. A useful safeguard should not only detect this risk during generation but also intervene while leaving safe responses largely unchanged and provide an auditable record of the internal patterns supporting its decisions. We propose TRACE (Temporal Risk Assessment and Concept Editing), a framework that tracks harmful-response risk during generation, selectively edits token activations, and leaves an auditable trace of its detection and steering decisions. TRACE learns a codebook of recurring activation patterns, with each codebook vector representing a latent concept. At inference time, it assigns each token activation to the nearest codebook vector to obtain the corresponding concept. A gated recurrent unit (GRU) processes the activations and their assigned concept features to predict completed-response harmfulness and a monotone cumulative streaming risk. During steering, activations assigned to harmful concepts are steered along directions defined by nearby benign targets. An external LLM labels the learned concepts post hoc, making detection sequences and harmful-to-benign edit records interpretable. Experiments across three generators and the WildGuard and S-Eval benchmarks show that TRACE improves streaming F1 over Kelp in five of six comparisons while remaining competitive on completed-response detection. TRACE's selective steering reduces Harmbench attack success with little change in refusal on originally safe responses. The post-hoc audit reveals recurring semantic structure in the learned concepts and distinct traces associated with harmful detection, safe generation, and activation edits.

    Fallback brief based on the available RSS feed; no unsupported details have been added.

    Source: arXiv
    Original article: https://arxiv.org/abs/2610.15365

  • RECAP: Calibrating Low-Bit Reasoning Models on the History They Read

    RECAP: Calibrating Low-Bit Reasoning Models on the History They Read

    arXiv:2610.15259v1 Announce Type: new Abstract: Low-bit quantization reduces the cost of serving reasoning models but can substantially degrade long chains of thought. Each step reads cached keys and values spanning thousands of tokens, allowing quantization errors to accumulate along the trajectory. Standard calibration operates on short segments that omit the earlier context needed by later reasoning steps. Even when this context is retained, cache reconstruction error alone does not capture how quantization errors affect subsequent attention. We present RECAP, a post-training quantization method that calibrates reasoning models on the history they read. Dynamic-prefix calibration separates long historical context from short-block reconstruction. It regenerates prefix keys and values with the current quantizer and rotates the active block to supervise every position along the trajectory. Prefix read geometry further supervises historical cache errors according to their effects on attention outputs, weighting key and value errors asymmetrically. With both components applied only during calibration, RECAP outperforms the baseline across multiple reasoning models and quantization settings. Under W4A4KV4, RECAP improves accuracy over the baseline by 13.3 percentage points on MATH500 with DeepSeek-R1-Distill-Qwen-1.5B and 15.8 points on AIME-120 with Qwen3-4B. These results highlight the importance of preserving historical dependencies during calibration for low-bit reasoning.

    Fallback brief based on the available RSS feed; no unsupported details have been added.

    Source: arXiv
    Original article: https://arxiv.org/abs/2610.15259

  • OntoAR-HAR: Ontology-Guided Agentic RAG with Personalized Memory for Human Activity Recognition

    OntoAR-HAR: Ontology-Guided Agentic RAG with Personalized Memory for Human Activity Recognition

    arXiv:2610.15292v1 Announce Type: new Abstract: LLMs increasingly support sensor-based human activity recognition (HAR) by linking motion signals with activity semantics and physical knowledge. Retrieval-augmented generation (RAG) grounds decisions in sensor examples and domain features, enabling training-free target-user inference with less dependence on raw numerical arrays. However, existing LLM/RAG-based HAR typically searches single-level activity spaces using static population knowledge. Consequently, they compare many unrelated candidates using the same static population evidence and overlook personalized motion patterns. We propose OntoAR-HAR, an ontology-guided agentic RAG framework combining static population and personalized knowledge. Its tree-structured activity ontology decomposes classification into local decisions among semantically related activities. The static store supplies node-constrained examples, while offline-discovered executable signal features provide numerical evidence for local distinctions. During inference, the personalized store collects selected user observations, which the LLM integrates with static knowledge and signal features while traversing the ontology to an activity label. Cross-subject and heterogeneous-sensing experiments show complementary gains from hierarchical retrieval and personalized memory across multiple LLM backbones, achieving state-of-the-art performance. By explicitly modeling hierarchical relations across activity granularities, OntoAR-HAR also addresses the long-overlooked semantic inclusion problem and avoids treating related labels as independent categories.

    Fallback brief based on the available RSS feed; no unsupported details have been added.

    Source: arXiv
    Original article: https://arxiv.org/abs/2610.15292