paperlist.ai

AI conference papers

AI conference paper library

Search papers from 30+ top conferences, including NeurIPS, ICLR, ICML, CVPR, ACL and AAAI. Filter by keyword, then open any paper for its abstract, authors and links.

arXiv

Real-Time Hard Peak Age-of-Information Safety with No-Regret Learning

Wentao Zhang, Wentao Mo

Safety-critical IoT systems such as industrial closed-loop control, V2X coordination, and remote teleoperation require every sensor's peak Age of Information (peak AoI, also abbreviated PAoI) to stay below a hard per-slot deadline, not merely an average bound. Existing approaches meet this requirement only under restrictive assumptions: stochastic channels for Whittle-index AoI, simulator rollouts for deep reinforcement learning, or sublinear cumulative violation for long-term constrained online convex optimization. Under adversarial coefficients, OCO-PAoI-Hard guarantees zero per-slot violation of the modeled AoI state under one-step viability and O(sqrt(T)) regret against any static safe comparator; packet-level safety requires stronger service assumptions. Our key observation is that the fractional peak-AoI deadline collapses exactly to an affine half-space constraint on the resource-allocation vector, turning hard real-time scheduling into time-varying constrained online convex optimization over a polyhedral safe set. A strictly causal proposal-shield-update loop enforces feasibility through one Euclidean projection per slot, the gradient step preserves no-regret behavior, and the classical virtual queue is reduced to an a-posteriori certificate. We establish closed-form static and dynamic regret bounds, a matching Omega(sqrt(T)) minimax lower bound, a margin-safe variant against execution noise, and a deadline-induced competitive ratio. On a four-sensor adversarial fluid-model trap channel, OCO-PAoI-Hard attains zero modeled-state deadline violations across all ten seeds, while four representative baselines miss between 1.65 percent and 64.0 percent of slots, and the empirical normalized regret stays below the theoretical envelope across two orders of magnitude in T.

arXiv

Reviewer Scores Are Not Comparable Across Research Areas in ML Peer Review

Binyan Xu, Xilin Dai, Fan Yang, Kehuan Zhang

Peer review at ML conferences increasingly relies on reviewer scores as the primary decision instrument. As submissions have scaled from thousands to tens of thousands per year, no systematic audit has examined whether this instrument functions uniformly across research areas, or whether acceptance outcomes are in practice shaped by forces that reviewer scores neither capture nor control. This position paper argues that acceptance outcomes are shaped by forces beyond reviewer scores, and that the underlying cause is a measurement design failure, not individual bias. When a fixed numerical scale aggregates quality judgments across communities with structurally non-uniform reviewer pools, absolute scores become incomparable across areas, and area chairs must substitute community priors for score-based decisions. Using ICLR 2021--2026 data covering 50,289 papers across 219 research topics, we show that at any given reviewer score, a paper's acceptance probability varies by up to 8x depending on its topic. We rule out scoring culture, expert reviewer standards, rational area chair reweighting, and quality dilution as alternative explanations. We call on program committees to adopt inherently calibrated review signals and publish topic-stratified, score-conditional acceptance rates as a first-class fairness metric.

arXiv

Early Failure Prediction from Near-Anomaly Detection: A Proactive Approach

L{é}a Billet, Louise Trav{é}-Massuy{è}s, Elodie Chanthery, Alexandre Gaffet

Anomaly detection methods often have uncertain behavior with respect to samples near the distribution boundary, limiting their ability to anticipate future anomalies. This work introduces the concept of near-anomalies that, while not yet anomalous, lie close to the boundary and are likely to transition into anomalies in the near future. To address this, we propose an unsupervised method, named Christoffel-based ANomaly Anticipation for eaRly dIscovery (CANARI), which leverages the strong theoretical foundations of the Christoffel function to detect near-anomalies. The method is validated on industrial in-circuit testing data from printed circuit boards, with synthetically generated near-anomaly samples due to the lack of real-world data labeling. Experimental results show that CANARI outperforms the compared baselines that generally use a dual-threshold mechanism (one for anomalies and one for near-anomalies). It therefore provides a proactive solution for anticipating anomalies before they occur, offering a promising approach for resilience, predictive maintenance, and quality control.

arXiv

Weak-to-Strong On-Policy Distillation

Fangxu Yu, Weijia Xu, Michael Xu, Tianyi Zhou, Zinan Lin

On-policy distillation (OPD), which aligns a student with the teacher's token-level distribution on the student's own rollouts, is an effective paradigm for transferring capabilities across LLMs. Prevailing approaches assume a teacher at least as capable as the student: they either distill a larger model into a smaller one, which fails at the frontier where no larger teacher exists, or consolidate multiple domain experts trained from a shared base, which requires costly training at the student's scale. We introduce Weak-to-Strong On-Policy Distillation (W2S-OPD), a simple yet effective OPD framework that improves the strong student by distilling from multiple weak models. W2S-OPD constructs a proxy teacher in logit space from a contrast pair of a positive and a negative model, both smaller than the student and cheap to obtain. Their logit difference isolates the capability direction, which is added to the student's own base model, yielding a proxy teacher that couples this direction while staying distributionally adjacent to the student. The student then distills it by minimizing the per-token reverse KL on its own rollouts. We instantiate the contrast pair as i) a post-RL expert against its pre-RL initialization, isolating the skill RL instills, ii) a larger against a smaller base model, isolating the capability from scale, and iii) a small base model with correct versus wrong hints, isolating the instance-level direction toward the solution. Across four math and three code benchmarks, W2S-OPD outperforms OPD, enables the student to surpass the domain teacher, and keeps improving the student even when every supervision source is weaker. Analysis shows different contrasts yield distinct signals: the post-RL and hint contrasts emphasize reasoning frameworks, while the scale contrast emphasizes the solving procedure. Our code will be available at https://github.com/Yu-Fangxu/W2S-OPD.

arXiv

SpecFormer: Mitigating Embedding and Attention Collapse via Spectral-Aware Transformer for Recommendation

Yu Cui, Yi Xu, Jiahao Wang, Hao Zhang, Yu Zhang, Xiaoyi Zeng, Can Wang, Jinxin Hu, Jiawei Chen

Transformer architectures have achieved remarkable success across diverse domains; however, directly applying their standard self-attention mechanism to recommendation often yields suboptimal performance, sometimes even trailing behind well-designed simple recommendation models. In this paper, we reveal that this performance bottleneck stems from severe embedding and attention collapse unique to recommendation scenarios. The heterogeneity and long-tail nature of recommendation data lead to a severe spectral collapse dominated by a few principal singular values. We further theoretically demonstrate that this triggers a vicious cycle in recommendation model's forward and backward propagation, which accelerates embedding and attention collapse and limits the model's scaling capability with increased depth. To address these issues, we propose SpecFormer, a novel Spectral-Aware Transformer designed for mitigating embedding and attention collapse in recommendation. Specifically, SpecFormer introduces 1) a Learnable Spectral Softening module to dynamically smooth the singular values distribution of the input token embeddings; 2) a Spectrum-softened Attention mechanism to model feature interaction under a more uniform spectral distribution space; 3) a Spectral Residual Position Encoding via Taylor expansion of singular values, explicitly providing a spectral inductive bias for feature interactions. Extensive experiments on one industrial and two public datasets demonstrate that SpecFormer significantly outperforms state-of-the-art baselines. Notably, SpecFormer has been successfully deployed in a real-world commercial recommender system and exhibits exceptional scaling capabilities: stacking SpecFormer layers actively improves the attention effective rank and recommendation performance.

arXiv

RobotDancing: Residual-Action Reinforcement Learning Enables Robust Long-Horizon Humanoid Motion Tracking

Zhenguo Sun, Yibo Peng, Yuan Meng, Xukun Li, Bo-Sheng Huang, Zhenshan Bing, Xinlong Wang, Alois Knoll

Long-horizon, high-dynamic motion tracking on humanoids remains brittle: retargeted reference motions are typically kinematically plausible but dynamically inconsistent with the robot, so small tracking errors accumulate and eventually destabilize control. We present RobotDancing, a practical single-stage reinforcement learning recipe that tracks retargeted motions by predicting reference-conditioned residual joint targets. By parameterizing actions as residual corrections on top of the reference, the policy can focus on compensating reference--robot dynamics mismatch (e.g., actuation limits, latency, friction, inertia) rather than re-synthesizing the motion. We train one policy per reference sequence while reusing a common training and deployment recipe across motions and platforms. We evaluate RobotDancing on eight LAFAN1 dance motions on Unitree G1 and conduct cross-platform experiments on H1 and H1-2. The resulting G1 policies execute long-horizon, high-energy behaviors on hardware without test-time modifications, and qualitative H1/H1-2 clips further illustrate cross-platform feasibility.

arXiv

A Rule-Based Approach to Specifying Preferences over Conflicting Facts and Querying Inconsistent Knowledge Bases

Meghyn Bienvenu, Camille Bourgaux, Katsumi Inoue, Robin Jean

Repair-based semantics have been extensively studied as a means of obtaining meaningful answers to queries posed over inconsistent knowledge bases (KBs). While several works have considered how to exploit a priority relation between facts to select optimal repairs, the question of how to specify such preferences remains largely unaddressed. This motivates us to introduce a declarative rule-based framework for specifying and computing a priority relation between conflicting facts. As the expressed preferences may contain undesirable cycles, we consider the problem of determining when a set of preference rules always yields an acyclic relation, and we also explore a pragmatic approach that extracts an acyclic relation by applying various cycle removal techniques. Towards an end-to-end system for querying inconsistent KBs, we present a preliminary implementation and experimental evaluation of the framework, which employs answer set programming to evaluate the preference rules, apply the desired cycle resolution techniques to obtain a priority relation, and answer queries under prioritized-repair semantics.

arXiv

ProbGuard: Proactive Runtime Monitoring for LLM Agent Safety via Probabilistic Prediction

Haoyu Wang, Christopher M. Poskitt, Jiali Wei, Jun Sun

Large Language Model (LLM) agents increasingly operate across domains such as robotics, virtual assistants, and web automation. However, their stochastic decision-making introduces safety risks that are difficult to anticipate during execution. Existing runtime monitoring frameworks, such as AgentSpec, primarily rely on reactive safety rules that detect violations only when unsafe behavior is imminent or has already occurred, limiting their ability to handle long-horizon dependencies. We present ProbGuard, a proactive runtime monitoring framework for LLM agents that anticipates safety violations through probabilistic risk prediction. ProbGuard abstracts agent executions into symbolic states and learns a Discrete-Time Markov Chain (DTMC) from execution traces to model behavioral dynamics. At runtime, the monitor estimates the probability that execution will remain safe from the current state, and triggers an intervention when this probability falls below a user-defined threshold. To improve robustness, ProbGuard incorporates semantic validity constraints in the abstraction and admits a PAC-style analysis that characterizes the sample complexity required to certify the learned model under standard assumptions. We evaluate ProbGuard in two safety-critical domains: autonomous driving and embodied household agents. Across evaluated scenarios, ProbGuard consistently predicts traffic law violations and collisions in advance, with warnings up to 15.84 seconds at a threshold yielding no false alarms, and up to 38.66 seconds at stricter thresholds. In embodied agent tasks, ProbGuard's re-prompting intervention mode reduces unsafe behavior by 65.37% relative to the unmonitored baseline while retaining 80.4% of the baseline task completion; a stricter halting configuration reduces unsafe behavior by 93.60% at a larger cost in completion.

arXiv

CEL: Comprehensive Counterfactual Explanations Library and Benchmark

Oleksii Furman, Łukasz Lenkiewicz, Marcel Musiałek, Maciej Zięba

Counterfactual explanations are a prominent approach in explainable artificial intelligence (xAI), providing actionable guidance on what input changes would alter a model's prediction to a desired outcome. While early methods primarily focused on minimal feature changes, recent work incorporates additional properties such as sparsity, actionability and plausibility. Despite this progress, fair and systematic evaluation remains challenging. Existing studies often rely on different data splits, predictive models, and evaluation metrics, which limits objective comparison across methods. To fill this gap, we introduce CEL (Counterfactual Explanations Library), a unified library and benchmark for counterfactual explanations designed to support consistent implementation and evaluation. CEL includes 18 datasets of varying size and complexity and provides implementations or reimplementations of 14 widely used counterfactual methods. Using this standardized setup, we conduct a comprehensive quantitative comparison across a variety of methods on datasets that differ in size, number, and types of attributes. The evaluation protocol incorporates multiple complementary metrics capturing validity, coverage, sparsity, proximity, and distributional plausibility, including density- and outlier-based measures to assess the realism of generated counterfactuals. To the best of our knowledge, this is the first comprehensive benchmark that systematically evaluates recent counterfactual explanation methods within a unified and reproducible framework. While prior libraries and benchmarking efforts exist in the literature, many are outdated, limited in scope, or lack consistent evaluation protocols. The proposed benchmark aims to improve reproducibility, enable fair comparison, and establish a workbench for the development of future counterfactual explanation methods.

arXiv

Physical Self-Supervised Learning: IMU Sensing without Manual Labels

Yuyang Leng, Renyuan Liu, Shaohan Hu, Peijun Zhao, Chun-Fu Chen, Songqing Chen, Shuochao Yao

Deep neural networks have become a promising approach for IMU-based sensing, but their scalability is fundamentally limited by costly labeled data and poor robustness to heterogeneous devices, placements, and users. Existing unsupervised and self-supervised methods reduce but do not remove this dependence, still requiring labeled data for domain adaptation and largely ignoring known physical structure. We propose physical self-supervised learning, an autoencoder-style paradigm for label-free IMU sensing. We replace the conventional neural decoder with an auto-adaptive physics decoder, a learnable family of kinematic equations that enforces explicit physical structure while adapting across environments, and adopt a hybrid two-stage IMU encoder with reconstruction in a structured latent space to mitigate sensor noise. Our framework further introduces probabilistic frequency-spatial constraints to disentangle sensor and object motion, a multi-view kinematic tree to exploit sparse physical self-supervised signals, and an uncertainty-aware formulation to handle the inherent ambiguity of IMU inference. Evaluated on inertial tracking and full-body motion capture over public datasets and realistic deployments, physical self-supervised learning reduces errors by up to 5x for tracking and 4x for motion capture in challenging generalization scenarios, consistently outperforming state-of-the-art supervised and self-supervised baselines without any labels. Our code is available at https://github.com/YuyangLeng/physical-ssl-imu-label-free

arXiv

CriPO: Enhancing Rubric-based RL via Self-Distillation

Mingxuan Xia, Yuhang Yang, Chao Ye, Shuai Zhu, Shenzhi Yang, Guangcheng Zhu, Yuhang Zhang, Cheng Peng, Haobo Wang, Siqing Wang

Rubric-based RL has recently shown promise in improving LLMs on open-ended tasks. A widely recognized limitation of rubric-based RL is limited exploration: criteria that no rollout manages to satisfy (Unexplored Criteria, UC) receive no optimization signal. Recent methods address this by incorporating rubric information as external guidance during rollout, yet they introduce a train-inference mismatch: the policy is optimized on rollouts produced under external guidance while this guidance is absent at inference time, causing error accumulation through autoregressive decoding. Moreover, these exploration-focused approaches overlook a fundamentally different failure mode that we term Suppressed Criteria (SC) -- criteria that are satisfied by some rollouts yet whose learning signals are lost during optimization because scalar reward aggregation assigns them non-positive aggregate advantages. Our analysis reveals that SC are remarkably prevalent: over 57% of samples exhibit this failure mode throughout training, with an average of 1.8 SC per sample. To simultaneously address both UC and SC without introducing training-inference mismatch, we propose Criterion-Distilled Policy Optimization (CriPO), which enhances rubric-based RL via on-policy self-distillation. For UC, CriPO constructs a criterion-injection self-teacher and computes a localized forward-KL loss to inject missing behaviors into the policy. For SC, CriPO employs a counterfactual self-teacher to locate criterion-relevant tokens in negative-advantage rollouts and flips their token-level advantages to positive values, preserving useful patterns that would otherwise be suppressed. Experiments on medicine and science benchmarks demonstrate that CriPO consistently outperforms rubric-based RL, achieving stronger final performance with approximately $2\times$ fewer optimization steps.

arXiv

DynImmune-BERT: Dynamic Immune Repertoire Modeling with Neural ODE Driven Continuous Transformers

Rong Fu, Yongtai Liu, Xiaowen Ma, Haoyu Zhao, Shuo Yin, Yiqing Lyu, Long Zhang, Wangyu Wu

Longitudinal T cell receptor repertoires contain signals of clonal expansion, contraction, disappearance, and reappearance after immune perturbation. Static repertoire language models usually summarize a sample as a bag of sequences, so the sampling interval, sequencing depth, and clone presence pattern are only weakly represented. This paper presents DynImmune-BERT, a continuous time repertoire model for patient level immune status prediction. The method combines depth adaptive centered log ratio initialization, clone presence gated Neural ordinary differential equation dynamics, bounded neighborhood self attention, event based state restart, and a hybrid transport objective that supervises dominant and rare clone mass. A low rank meta adapter initializes reappearing clonotypes while keeping the parameter count independent of the number of observed clones. The evaluation separates literature reported baselines from internally controlled temporal comparisons, reports uncertainty for small external cohorts, adds calibration and threshold diagnostics, and visualizes latent clone trajectories and attention neighborhoods. The results indicate that event aware temporal modeling can complement strong static encoders when longitudinal repertoire structure is available, while small external cohorts and protocol differences require cautious interpretation.

arXiv

Rethinking Group Recommender Systems in the Era of Generative AI: From One-Shot Recommendations to Agentic Group Decision Support

Dietmar Jannach, Amra Delić, Francesco Ricci, Markus Zanker

More than twenty-five years ago, first ideas were developed on how to design a system that can provide recommendations to groups of users instead of individual users. Since then, a rich variety of algorithmic proposals were published, e.g., on how to acquire individual preferences, how to aggregate them, and how to generate recommendations for groups of users. However, despite the rich literature on the topic, barely any examples of real-world group recommender systems can be found. This lets us question common assumptions in academic research, in particular regarding communication processes in a group and how recommendation-supported decisions are made. In this essay, we argue that these common assumptions and corresponding system designs often may not match the needs or expectations of users. We thus call for a reorientation in this research area, leveraging the capabilities of modern Generative AI assistants like ChatGPT. Specifically, as one promising future direction, we envision group recommender systems to be systems where human group members interact in a chat and an AI-based group recommendation agent assists the decision-making process in an agentic way. Ultimately, this shall lead to a more natural group decision-making environment and finally to wider adoption of group recommendation systems in practice.

arXiv

From Preimage Search To Source-Grounded Feature Inversion

Kaixiang Shu

Interpreting a neural network requires understanding what its internal features extract from a particular input. Feature inversion seeks to express a selected feature in the input domain, but canonical iterative methods search for an input whose re-encoded representation matches the target. Because many inputs can satisfy this constraint, target matching alone does not specify the inverse associated with the sample that generated the feature. We formulate source-grounded feature inversion by conditioning the inverse on the source-local network geometry at the target-generating input. At each boundary of the computational DAG, backpropagation provides the correct reverse dependencies but transports an adjoint signal rather than an upstream-state estimate. We locally repair this signal with a closed-form matrix Wiener map from a mean-seed VJP to the upstream state, followed by a second Wiener map for the JVP forward-consistency residual, and compose the repaired states through the same DAG in one finite reverse pass. One calibrated zero-intercept map family supports new inputs, depths, channels, and channel groups across diverse CNN and Transformer architectures, tensor components, and visual distributions without query-specific optimisation. Matched target and source controls verify that each inverse depends on the selected feature and the local operators of the sample being explained, rather than a target-independent image template. Prediction-conditioned feature atlases align these visualisations with independent interventions on the corresponding internal features. Together, source-grounded feature inversion opens the model's hidden feature hierarchy to inspection at the level of individual layers and channels, linking what the network extracts from an input to the internal evidence that shapes its decision.

arXiv

Grounded Vision-Language Interpreter for Long-Horizon Bimanual Task and Motion Planning

Jeremy Siburian, Keisuke Shirai, Cristian C. Beltran-Hernandez, Masashi Hamaya, Michael Görner, Atsushi Hashimoto

While recent advances in vision-language models have accelerated language-guided robot planning, their black-box nature lacks the safety guarantees and interpretability crucial for real-world deployment. Conversely, classical symbolic planners offer rigorous safety verification but require significant expert knowledge for setup. Moreover, most existing methods are limited to single-arm pick-and-place tasks, leaving bimanual manipulation largely underexplored, despite its tightly interdependent subtasks and the need to explicitly manage inter-arm collisions. To bridge this gap, this paper proposes ViLaIn-TAMP, a hybrid planning framework for enabling verifiable, interpretable, and autonomous bimanual robot behaviors. ViLaIn-TAMP comprises three main components: (1) a Vision-Language Interpreter (ViLaIn) adapted from a prior work that converts multimodal inputs into structured PDDL problem specifications, (2) an integrated Task and Motion Planning (TAMP) system that grounds these specifications in actionable trajectory sequences through symbolic and geometric constraint reasoning, explicitly verifying feasibility before execution, and (3) a corrective planning (CP) module which receives structured motion failure feedback and feeds it as constraints back to ViLaIn to refine the specification. We design challenging bimanual manipulation tasks in a cooking domain to evaluate our framework, where experimental results show that ViLaIn-TAMP outperforms a VLM-as-a-planner baseline by 17.5% in mean success rate, with the CP module boosting it further by 32.9%. We further validate ViLaIn-TAMP on a physical dual-arm robotic system. Project page: https://omron-sinicx.github.io/ViLaIn-TAMP

arXiv

MuScriptor: An Open Model for Multi-Instrument Music Transcription

Simon Rouard, Michael Krause, Axel Roebel, Carl-Johann Simon-Gabriel, Alexandre Défossez

Existing methods for automatic music transcription are often limited to single-instrument recordings or fail on complex, real music mixes. Although previous work utilizes synthetic training data, the resulting models generalize poorly, leading to largely unusable transcription output in realistic, multi-instrument settings. In this work, we analyze the effectiveness of synthetic data for pre-training while combining it with fine-tuning on real music audio and post-training using reinforcement learning. We further introduce conditioning on instrument presence to customize transcriptions. Finally, we release MuScriptor, an open-weight multi-instrument music transcription model that works on real-world music recordings from across a diverse range of musical genres.

arXiv

ECHO: Prune To Act, Trace To Learn With Selective Turn Memory In Agentic RL

Zijun Xie, Binbin Zheng, Enlei Gong, Jihua Liu, Yuyang You, Lingfeng Liu, Jiayao Tang, guanqun zhao, Aoqi Hu, Zeyu Chen

Long-horizon language agents must repeatedly interact with tools, accumulate evidence, and make decisions under bounded context windows. Context-management methods make such rollouts feasible by simplifying past interactions through deletion, folding, or memory editing. However, when useful history is collapsed into compressed states, the reconstructed context may no longer reveal which earlier observations support a successful final answer. This creates a mismatch between bounded-context acting and outcome-based reinforcement learning: the policy acts on reconstructed context, while the learner lacks source-level provenance for assigning credit to the evidence that mattered. We propose ECHO, a selective turn-memory framework for traceable context reconstruction in Agentic RL. ECHO compresses each completed environment turn into a compact source-indexed memory record, reconstructs bounded policy contexts by selecting useful records, and reuses the selected source indices to route positive outcome credit to the final trajectory segment, reused evidence turns, memory findings, and memory-selection actions. On BrowseComp-Plus, ECHO reaches 43.4% held-out accuracy, outperforming GRPO at 28.9% and the rolling-summary baseline SUPO at 36.1%, while using fewer turns and lower trajectory volume than SUPO. The trained policy also improves zero-shot generalization across multi-objective QA, code generation, and deep information-seeking benchmarks on both dense and MoE backbones.

arXiv

Prefix-Guided On-Policy Distillation: Mining Golden Trajectories from Rollouts

Qingfei Zhao, Huan Song, Shuyu Tian, Jiawei Shao, Xuelong Li

On-policy distillation (OPD) improves reasoning models by applying dense teacher supervision on student-sampled trajectories. However, scaling OPD to long-horizon reasoning exposes a reliability and efficiency problem: standard OPD assigns every candidate the same long rollout budget, even though some trajectories may quickly become weakly aligned with the teacher and provide less useful supervision. Prior analyses suggest that teacher--student compatibility is important for OPD success, motivating early-prefix top-k overlap as a proxy for continuation value. Continuing low-overlap trajectories to the full rollout length may cause them to drift further from the teacher as generation proceeds, increasing computational cost while providing limited distillation benefit. To address this, we introduce Prefix-Guided On-Policy Distillation (PG-OPD), a rollout-allocation framework that estimates continuation value from fixed-length prefixes. PG-OPD computes teacher-student top-k overlap in an early probe window and allocates long rollouts only to high-overlap candidates, while stopping the rest at the prefix length. Across teacher--student combinations on AMC, AIME, and HMMT benchmarks, PG-OPD achieves up to a 4.80-point accuracy gain and up to a 2.46x wall-clock speedup across configurations. These results demonstrate that using early-prefix compatibility to guide candidate pruning can improve both training efficiency and reasoning performance.

arXiv

HERALD: High-Throughput Block Diffusion LLM Serving via CPU-GPU Cooperative KV Cache Retrieval

Omin Kwon, Doyeon Kim, Jongseok Park, Seung Yul Lee, Ion Stoica, Jae W. Lee

The KV cache dominates GPU memory in long-context LLM serving, crowding out batch capacity and leaving GPU compute idle. Offloading the cache to CPU DRAM restores capacity, but the limited PCIe bandwidth forces state-of-the-art offloading systems to pair it with sparse attention, fetching only a small critical subset of the cache to the GPU. These systems, however, follow the KV access pattern of autoregressive decoding, in which the critical set changes at every token: selection and fetching recur at every decoding step, and throughput remains capped by PCIe bandwidth rather than by either processor. Block diffusion LLMs(block dLLMs), which decode a block of B tokens over T denoising steps, exhibit a different KV access pattern that opens a new opportunity for offloading. Recent sparse block dLLM methods have shown that sparse inference separates into a selection phase that scans the full KV cache once per block and a denoising phase that reuses the selected small subset T times. This asymmetry aligns with the compute and memory asymmetry of a CPU-GPU system, making it advantageous to run selection on the CPU and denoising on the GPU: the critical KV cache then crosses PCIe only once per block, removing the interconnect as the bottleneck. We present HERALD, to our knowledge the first KV offloading system designed for block dLLMs. HERALD resolves the two obstacles of this mapping, the serialized dependency between the phases and the compute-bound B-query selection on the CPU, by overlapping the phases with a draft block, reducing the selection cost with a single [MASK] query, and executing both as a dual-stream pipeline over double-buffered sparse KV pools. On two production block dLLMs, HERALD sustains near-lossless accuracy at a 5% KV budget and reaches up to 2.28x the decode throughput of GPU-only serving, with gains that widen with context length.

arXiv

Cost-Based Semantics for Querying Inconsistent Weighted Knowledge Bases

Meghyn Bienvenu, Camille Bourgaux, Robin Jean

In this paper, we explore a quantitative approach to querying inconsistent description logic knowledge bases. We consider weighted knowledge bases in which both axioms and assertions have (possibly infinite) weights, which are used to assign a cost to each interpretation based upon the axioms and assertions it violates. Two notions of certain and possible answer are defined by either considering interpretations whose cost does not exceed a given bound or restricting attention to optimal-cost interpretations. Our main contribution is a comprehensive analysis of the combined and data complexity of bounded cost satisfiability and certain and possible answer recognition, for description logics between ELbot and ALCO.

Keep browsing

More papers from top AI conferences

30+ top conferences, by conference, year, and topic.