跳到正文
研踪PaperTrail

论文库

检索计算机领域论文,按研究方向、作者或关键词筛选。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+1284新增论文
+1011其中 AI 相关
+150已有论文修订
+97新增 AI 导读

arXiv 最近完成扫描:2026-11-20 08:48 · 全站最近入库:2026-11-20 10:50各信源进度 →

统计口径与数据新鲜度

统计 2026-11-19 12:06 — 2026-11-20 12:06(北京时间) · 点击数字查看对应简报;AI 相关包含在新增论文内。

最近完成检查:2026-11-20 08:48最近发现新增版本:2026-11-20 10:50库内论文最新更新时间:2026-11-19 12:39

每分钟刷新已保存的数据。空检查不代表采集异常;新收录可能包含历史论文。AI 导读按新增版本计数。

28,155 已收录论文23,386 AI 相关8,624 有修订112 HF 社区发现

浏览论文库

时间范围跟随排序:版本更新、本站收录或首次发表。最新收录可能包含历史论文。部分信源异常,查看状态 →

找到 17 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-20 08:46(北京时间)
版本更新 · v2arXiv:2602.02227

Show, Don't Tell: Morphing Latent Reasoning into Image Generation

Harold Haodong Chen, Xinxiang Yin, Wen-Jie Shu, Hongfei Zhang, Zixin Zhang, Chenfei Liao, Litao Guo, Qifeng Chen 等 9 位作者

原文摘要

Text-to-image (T2I) generation has achieved remarkable progress, yet existing methods often lack the ability to dynamically reason and refine during generation--a hallmark of human creativity. Current reasoning-augmented paradigms most rely on explicit thought processes, where intermediate reasoning is decoded into discrete text at fixed steps with frequent image decoding and re-encoding, leading to inefficiencies, information loss, and cognitive mismatches. To bridge this gap, we introduce LatentMorph, a novel framework that seamlessly integrates implicit latent reasoning into the T2I generation process. At its core, LatentMorph introduces four lightweight components: (i) a condenser for summarizing intermediate generation states into compact visual memory, (ii) a translator for converting latent thoughts into actionable guidance, (iii) a shaper for dynamically steering next image token predictions, and (iv) an RL-trained invoker for adaptively determining when to invoke reasoning. By performing reasoning entirely in continuous latent spaces, LatentMorph avoids the bottlenecks of explicit reasoning and enables more adaptive self-refinement. Extensive experiments demonstrate that LatentMorph (I) enhances the base model Janus-Pro by $16\%$ on GenEval and $25\%$ on T2I-CompBench; (II) outperforms explicit paradigms (e.g., TwiG) by $15\%$ and $11\%$ on abstract reasoning tasks like WISE and IPV-Txt, (III) while reducing inference time by $44\%$ and token consumption by $51\%$; and (IV) exhibits $71\%$ cognitive alignment with human intuition on reasoning invocation.

cs.CV

版本更新 · v2arXiv:2611.23555

MetaphIR: Corpus-Scale Applicability Analysis for LLVM Peephole Optimizations

Hongxu Xu, Xintong Zhou, Chunhao Liao, Chengnian Sun

原文摘要

Peephole optimizations replace short source instruction sequences with more efficient equivalents, typically guarded by preconditions such as requiring a constant operand to be a power of two. Since every new rewrite adds maintenance burden and compile-time cost, compiler developers need to know how often real-world programs both match the source pattern and satisfy its preconditions before committing to an implementation. Existing workflows fall short: structural search over an intermediate representation (IR) corpus of 4.8 million functions ignores preconditions and therefore overcounts, whereas implementing the rewrite and recompiling the corpus yields exact counts only after paying the engineering cost, with each run of the corpus's on-demand recompilation pipeline taking a median of 47 minutes. This delay is especially limiting when generalizing a rewrite, as developers often iterate over multiple variants to trade off coverage against complexity. We introduce MetaphIR, a system pairing a domain-specific language for expressing LLVM peephole optimizations with an indexed search engine over large IR corpora. Using a syntax modeled on LLVM's C++ matchers, MetaphIR patterns augment LLVM IR with holes and Boolean preconditions over matched values. To answer queries at interactive speed, MetaphIR builds compact per-function summaries ahead of time and uses abstract interpretation over those summaries to soundly filter out functions that cannot satisfy a query's constraints. Across twenty recent peephole optimizations from LLVM's InstCombine pass, MetaphIR recovers the exact set of functions where each rewrite applies in isolation, achieving 100% precision and recall. Queries complete in a median of 1.27 seconds---a 22x speedup over a full corpus scan---enabling developers to quickly gauge an optimization's reach and generality prior to implementation.

cs.PL · cs.SE

版本更新 · v2arXiv:2611.23803

CoT-WAM: Collaborative Training of Vision-Language Models and World-Action Models via Plan Latents

Jinhao Zhang, Yushen Zuo, Ke Rui, Weitao Zhou, Haibao Yu, Honglong Tian, Xiaoheng Ma, Jie Mei 等 10 位作者

原文摘要

General robot manipulation requires both semantic competence and physical foresight. World action models (WAMs) acquire an understanding of and predictive capability over the physical world by jointly modeling visual foresight and action distributions. However, existing WAMs still lack semantic competence: they have limited ability to perform open-world scene understanding, semantic reasoning, and long-horizon planning. Although vision-language models (VLMs) provide these capabilities, effectively integrating their semantic priors into WAM prediction and control remains challenging. To address this challenge, we propose CoT-WAM, a jointly trained WAM-VLM framework. CoT-WAM uses a learnable embedding to exchange information between the VLM and the WAM, which is supervised by textual and goal-image signals during training to encode rich intent information. A tailored two-stage training strategy enables joint optimization and more effective collaboration between the VLM and WAM. We evaluate CoT-WAM on six benchmarks covering control performance, zero-shot generalization, and long-horizon planning. CoT-WAM achieves SOTA performance on LIBERO, CALVIN, and RoboCasa-Kitchen, and demonstrates stronger zero-shot generalization on LIBERO-Plus and LIBERO-Pro without fine-tuning. On RoboCasa365, it further improves long-horizon planning over baselines with and without external planners. Real-world experiments further demonstrate the effectiveness of our method in practical deployments. Code is available at: https://github.com/jhz1192/CoT-WAM.git

cs.RO

版本更新 · v3arXiv:2609.27964

Learning from Transients: Linear RNN Scaling Laws in Stationary Prediction

Ziyan Chen, Linhe Zheng, Peilin Liu, Zhongzhu Zhou, Ding-Xuan Zhou

原文摘要

Does the organization of a fixed number of transitions into independent trajectories affect a fixed stationary prediction task? We study a Gaussian-sketched linear recurrent teacher--student model, where independently initialized training trajectories create transients that can excite weak spectral directions. For full-batch gradient descent with stable deterministic learning rates, we establish matching stationary approximation, bias, variance, and total-risk scaling laws from dependent trajectory data. When initialization decays at least as fast spectrally as the innovations, trajectory count and length enter the leading variance rate through their product. With a heavier initialization tail, transient excitation can improve the trajectory-length-dependent finite-time bias and, at fixed transition count, variance, while the stationary oracle shift remains controlled by representation approximation. Under token and compute caps, constant-length trajectories attain the optimal risk scaling order within the theorem-controlled range. Synthetic experiments provide finite-scale evidence in this stylized model.

cs.LG

版本更新 · v3arXiv:2609.27365

Anchor and Perturb: Lazy Agent Remediation by Exploration Injection

Chengxi Zhong, Yongzhe Chang

原文摘要

Anchor and Perturb (AnP) is a lightweight framework that resolves multi-agent coordination failures by decoupling exploratory variance injection from recurrent manifold stability. Existing remediation strategies predominantly alter mixing network architectures or enforce simultaneous exploration across the collective, which inevitably precipitates severe temporal-difference penalties in non-monotonic reward spaces. Specifically, AnP isolates underperforming lazy agents and injects an asymmetric exploratory pulse into targeted coordinates whilst anchoring converged teammates to nominal greedy exploitation. Empirical telemetry benchmarks demonstrate that AnP successfully rescues collapsed joint policies (recovering from a 5% evaluation win rate nadir back to 85%) and facilitates escape from suboptimal coordination plateaus, sustaining peak win rates of 90% without requiring structural network modifications.

cs.MA

版本更新 · v4arXiv:2410.07468

Fast Mixed-Precision Real Evaluation

Artem Yadrov, Aditya Kumar, Pavel Panchekha

原文摘要

Evaluating real numbers to high precision is important in computational science, numerical computing, and in hardware and software design. Efficient algorithms exist for computing arithmetic operations and common elementary functions, but composing these primitives into larger expressions remains a challenge. The typical expression-evaluation algorithm computes all primitives in the expression to a uniform precision, doubling that precision until a sufficiently-accurate result is achieved. This is wasteful: usually only a few operations really need to be performed at high precision, and the bulk of the expression could be computed at a much lower, faster precision. However, no rigorous algorithm currently exists to compute these non-uniform precision assignments, nor to leverage non-uniform precision to reduce the number of primitive computations needed to evaluate an expression. We propose a new algorithm for deriving such rigorous precision assignments and reducing the number of primitive computations needed. This algorithm can leverage results computed at lower precisions to analytically determine a precision for each primitive that will result in a sufficiently-accurate evaluation. It can also use information from constant subexpressions, movability flags, and repeat flags to skip the computation of certain primitives, further reducing the time to evaluate real expressions. Our implementation, Reval, achieves an average speed-up of 3.97x compared to the widely-used Sollya tool, with the speed-up increasing to 9.26x on the most difficult inputs. A detailed evaluation shows that this speed-up results from both assigning lower precisions to the majority of primitives and from computing fewer primitives overall.

cs.MS · math.NA

新论文 · v1arXiv:2611.25489

When Should a Voice Agent Speak? End-of-Turn Detection in Turkish

Ahmet Tuğrul Bayrak, Fatma Nur Korkmaz, Mustafa Sertaç Türkel, Nur Yağmur Tuncer

原文摘要

A voice agent that waits for a fixed silence before replying either interrupts users who pause mid-utterance or responds late to users who have finished. This study examines the decision at the onset of silence, using turn ends and turn-internal pauses extracted from the Real-TurnTurk corpus of Turkish dyadic video calls. A fine-tuned Turkish TurnGPT model, a zero-shot Turkish language model and a deployed multilingual detector are compared, together with voice-activity timing and facial features, under recording-disjoint and speaker-disjoint evaluation with paired cluster bootstrap tests. On held-out recordings, the fine-tuned model separates turn ends from pauses better than both text baselines and is the best calibrated of the three. Timing features of the current speaker are similarly informative, and their combination with the fine-tuned text score gives the best timing-based results; with logistic regression, this combination also outperforms the same combination with either text baseline. In a latency simulation, adding the text score to a silence timeout shortens the mean response delay at the same interruption rate. On automatic Whisper transcripts, the fine-tuned model retains its advantage over the zero-shot model.

cs.AI · cs.CL

版本更新 · v2arXiv:2610.07593

TRANSIT: Transparent Scale-in for Multi-Node LLM Training

Hyungyo Kim, Nicholas Satchanov, Hrishi Shah, Gaohan Ye, Jiaqi Lou, Robert Walkup, Shweta Salaria, I-Hsin Chung 等 12 位作者

原文摘要

The growing memory requirements of large language models (LLMs) increasingly force distributed training jobs to span multiple GPU nodes to accommodate their training state. This reduces per-GPU computational efficiency, increases network traffic, and exacerbates job queueing delays. To address these inefficiencies, we introduce TRANSIT, a system that transparently enables efficient distributed LLM training on fewer GPUs through memory oversubscription. TRANSIT optimizes CPU-GPU data movement during oversubscription by coordinating driver-managed page residency with execution-guided zero-copy access, exploiting recurring kernel execution patterns across training iterations. We evaluate TRANSIT on dense and mixture-of-experts (MoE) models across different parallelization strategies using up to 64 NVIDIA H100-80GB GPUs interconnected over RDMA over Converged Ethernet (RoCE). Compared with the minimum GPU count required to accommodate the training state in GPU memory, TRANSIT enables training on up to 50% fewer GPUs while retaining over 90% of the baseline per-GPU computational throughput. In some configurations, training on 17-33% fewer GPUs even improves throughput due to reduced distributed communication overhead. When training on fewer GPUs, TRANSIT achieves up to 243% and 87% higher throughput than conventional driver-runtime-managed memory oversubscription and framework-managed CPU offloading, respectively, while reducing CPU memory consumption by up to 79% compared with the latter. Training on fewer GPUs with TRANSIT also reduces per-node network traffic by up to 33% and improves throughput by up to 44% under constrained network bandwidth. Production-trace-driven simulations further show 15-49% lower inter-rack network traffic and up to 2-95% shorter job queue times for different cluster capacities.

cs.DC · cs.AR

版本更新 · v2arXiv:2608.07981

Distilling Physical Priors into Streaming World Models

Liangliang Zhao, Junying Wang, Danni Yang, Yifan Chang, Bin Fu, Yu Qiao, Bowen Zhou, Yihao Liu

原文摘要

Streaming world models predict future visual states online while maintaining physically coherent dynamics over long horizons. However, their rollouts often violate basic physical constraints. A common approach distills pretrained bidirectional DiTs into few-step causal generators. However, this paradigm suffers from two fundamental limitations: generic bidirectional teachers acquire limited physical priors from visually oriented pretraining, and the limited priors suffer further loss during bidirectional-to-causal distillation. We present PhyS, a three-stage framework for distilling physical priors into streaming world models. To acquire physical priors from real-world interactions, we construct PhyS-120K, a dataset of 120K real-world physical-interaction videos spanning rigid-body dynamics, soft-body deformation, fluid phenomena, and phase transitions. Each video is annotated with structured descriptions of object properties and causal state transitions. Physics-aware supervised fine-tuning injects the physical priors into a bidirectional 14B DiT teacher, which we then distill into a lightweight 1.3B causal DiT for few-step autoregressive streaming generation. Finally, we use online reinforcement learning to incentivize the distilled model to generate physically plausible rollouts and further propose Temporal Credit Routing (TCR) to address temporal credit assignment. TCR evaluates physical consistency over overlapping temporal windows and routes the resulting group-relative advantages to temporally aligned denoising actions. On PhysicsIQ, PhyS improves the Wan2.1-14B teacher by 18.2\% and the Self Forcing, Rolling Forcing, and Causal Forcing by 23.7\%, 14.8\%, and 31.4\%, respectively. Results also improve the physics-aware video benchmarks VideoPhy, VideoPhy2, and PhyGenBench. The dataset, code, and more sample videos are available on our Project Page.

cs.CV

版本更新 · v2arXiv:2611.16413

RICapBench: A Fine-Grained Visual Understanding Benchmark Dataset for Vision-Language Models in Remote Sensing

Xue Wang, Yubin Xu, Kun Tan

原文摘要

Advances in vision-language models (VLMs) have improved remote sensing image analysis and supported geospatial decision-making. However, fine-grained understanding of remote sensing imagery remains challenging for VLMs. Large variations in object scale, complex geospatial relations among instances, and limited visual detail hinder precise instance localization, visible attribute identification, and spatial relation understanding, potentially leading to descriptions that lack visual grounding. Existing datasets and benchmarks insufficiently assess end-to-end instance localization and captioning without target-specific textual cues. To support this evaluation, we introduce RICapBench, a dataset and benchmark that unifies instance localization and fine-grained captioning in remote sensing. RICapBench contains 245,466 instance-level annotations and 498,005 spatial relation tuples spanning general remote sensing, UAV remote sensing, and dense small-object scenarios. Each annotation pairs an object's oriented bounding box with a natural language description of its visible attributes and spatial relations to surrounding instances and environmental elements. To evaluate the accuracy and completeness of spatial relation captioning, we propose a remote sensing spatial relation parser and the RelCap-F1 metric. We combine RelCap-F1 with instance localization and attribute captioning metrics into a composite metric, RICScore. Our evaluation of 15 advanced open-source and closed-source VLMs identifies instance localization as a major bottleneck for these models. LocateAnything achieves the highest RICScore of 31.68, but scores lower on spatial relation captioning than other task-fine-tuned open-source VLMs. Code and dataset are available at https://github.com/Xldln/RICapBench.

cs.CV

版本更新 · v2arXiv:2610.34730

The rank of $\mathbb{Z}_{2^s}$-linear Hadamard codes

Xuan Wang, Minjia Shi

原文摘要

The canonical $\mathbb{Z}_{2^s}$-linear Hadamard codes are the Gray images of the $\mathbb{Z}_{2^s}$-additive codes. Their rank, the dimension of the binary linear span, was known in closed form only for $s\le 3$, and for $s \ge 4$ only through computer computations for small lengths. We give a method that computes the rank for every $s$. We show that the rank is a binomial polynomial in the type parameters, with coefficients $ρ_s(\boldsymbol{k})$ independent of the length. We compute all the coefficients for $s\le 5$, which yields explicit rank formulas over $\mathbb{Z}_4$ and $\mathbb{Z}_8$ (recovering the known ones), and over $\mathbb{Z}_{16}$ and $\mathbb{Z}_{32}$.

cs.IT

版本更新 · v2arXiv:2611.24455

Homeward: No Place Like the Training View

Mingyang Lyu, Yinqian Sun, Gongce Cao, Guobin Shen, Yuhao Zhang, Lu Jia, Chenfei Yan, Qian Zhang 等 11 位作者

原文摘要

Vision-language-action (VLA) policies implicitly assume that third-person cameras remain fixed at their data-collection poses; once a camera is bumped, remounted, or swapped at deployment, task success degrades precipitously even though the underlying task and robot remain unchanged. We present Homeward, a training-free, plug-and-play observation adapter that canonicalizes observations from an uncalibrated deployment camera back into the training view for frozen VLA policies. Our design stems from a closed-loop diagnostic paradox in simulation: while re-rendering perturbed views into the training camera with oracle geometry recovers much of the lost performance, substituting an off-the-shelf feed-forward 3D foundation model can leave the policy worse off than applying no correction at all. This degradation concentrates on the thin, articulated robot arm, which feed-forward depth prediction tears and scatters - yet the arm's exact 3D state is already known from robot kinematics. Homeward exploits this asymmetry: it estimates what is unknown - camera parameters and static scene depth - from the live RGB frame and a single training-view reference, and renders what is known - the moving robot links - directly from the kinematic model. Evaluated across $\sim$300,000 closed-loop episodes spanning four frozen VLAs, Homeward boosts success across all 1,599 camera variants of LIBERO-Plus from 62.5%, 76.5%, 68.0%, and 3.6% to 77.6%, 95.1%, 91.2%, and 26.4% for OpenVLA-OFT, $π_{0.5}$, $π_0$, and OpenVLA, respectively. Without any scene fine-tuning or deployment camera calibration, Homeward extends to 6-DoF random placements, focal-length changes, and, with per-step pose estimation, continuous intra-episode camera drift, while matching or surpassing a calibrated, scene-trained view-synthesis baseline on three of four policies on LIBERO-Plus.

cs.RO

版本更新 · v2arXiv:2403.07518

Open-Vocabulary Scene Text Recognition via Pseudo-Image Labeling and Margin Loss

Xuhua Ren, Hengcan Shi, Jin Li

原文摘要

Scene text recognition is an important and challenging task in computer vision. However, most prior works focus on recognizing pre-defined words, while there are various out-of-vocabulary (OOV) words in real-world applications. In this paper, we propose a novel open-vocabulary text recognition framework, Pseudo-OCR, to recognize OOV words. The key challenge in this task is the lack of OOV training data. To solve this problem, we first propose a pseudo label generation module that leverages character detection and image inpainting to produce substantial pseudo OOV training data from real-world images. Unlike previous synthetic data, our pseudo OOV data contains real characters and backgrounds to simulate real-world applications. Secondly, to reduce noises in pseudo data, we present a semantic checking mechanism to filter semantically meaningful data. Thirdly, we introduce a quality-aware margin loss to boost the training with pseudo data. Our loss includes a margin-based part to enhance the classification ability, and a quality-aware part to penalize low-quality samples in both real and pseudo data. Extensive experiments demonstrate that our approach outperforms the state-of-the-art on eight datasets and achieves the first rank in the ICDAR2022 challenge.

cs.CV

版本更新 · v3arXiv:2610.28480Berkeley BAIR

SuperMimic: Scaling Humanoid Whole-Body Control with One Differentiable World Model

Qiayuan Liao, Yiyang Shao, Junfeng Long, Kevin Zakka, Roberto Horowitz, Pieter Abbeel, Koushil Sreenath

原文摘要

Tracking large human-motion datasets has become the dominant route to general, expressive humanoid whole-body control, but scaling across many motions is expensive. Current pipelines use model-free reinforcement learning, either merging per-motion specialists through lossy distillation, or brute-forcing a single policy with cluster-scale compute. SuperMimic amortizes learning across all motions with one differentiable world model of their shared dynamics, fit from logged transitions of a standard rigid-body simulator. The world model and the tracking policy train together in one loop, and backpropagation-through-time over short imagined rollouts gives the policy analytic gradients. Fast imagined rollouts and analytic gradients make policy learning efficient. In 3 hours on a single GPU, SuperMimic surpasses the return of a state-of-the-art motion tracker baseline trained for 7 days on 128 GPUs. It reduces GPU-hours by more than $1000\times$ and fails on only 25 of 129,514 benchmark motions, versus 179 failures for the baseline. After 9.8 hours on two GPUs, SuperMimic achieves zero failures. From 1 to 264 hours of motion, the world model's prediction error stays stable, and GPU-hours to a target return grow slowly. Our policy transfers to a physical Unitree G1 across diverse and acrobatic motions. SuperMimic removes compute as the bottleneck to scaling humanoid whole-body control, making it practical to measure motion-tracking scaling laws.

cs.RO

新论文 · v1arXiv:2611.25480

Learning Ankle Exoskeleton Control for Reactive Standing Balance Through Neuromusculoskeletal Simulation

Maria T. Tagliaferri, Inseung Kang

原文摘要

Deficits in reactive balance can increase fall risk among older adults and individuals with neuromuscular impairments, motivating the need for interventions that improve balance recovery. Wearable robotic exoskeletons offer a promising approach to addressing this need by augmenting balance control. However, developing controllers that deliver appropriately timed and scaled assistance remains challenging, as evaluating each candidate strategy requires repeated exposure to destabilizing perturbations. We propose a simulation-to-real framework that combines motion imitation and reinforcement learning to develop balance-assistive ankle exoskeleton control. A muscle-driven neuromusculoskeletal model was trained using a small set of motion-capture data from a single subject to reproduce kinematic and kinetic responses to backward-translation perturbations. This model served as a virtual human for training a reinforcement-learning exoskeleton controller. The resulting controller generated bilateral ankle assistance using only inertial sensor measurements, without explicit perturbation detection. We implemented the controller on-board a bilateral ankle exoskeleton and evaluated it in five healthy subjects whose data were not used for training. Assistance increased the perturbation magnitude that subjects could withstand without stepping by 11.8% and reduced peak center-of-mass displacement and velocity by 13.5% and 28.2%, respectively. Individual benefits varied with baseline recovery strategy and body mass. These results indicate that a single virtual human can yield transferable balance assistance, and that more diverse virtual humans could further improve its effectiveness.

cs.RO

版本更新 · v2arXiv:2611.24451

EviScope: Evidence-Conditioned Sampling for Long-Range 3D Detection

Mingyang Lyu, Yinqian Sun, Gongce Cao, Chujie Tang, Yuhao Zhang, Lu Jia, Qian Zhang, Bo Wang 等 9 位作者

原文摘要

Query-based camera-radar detectors choose where and over which frames each object query samples from the query's position or feature. We test whether radar evidence around the query should also set these choices at the 150 m range of MAN TruckScenes, where recall differs across object-level radar-return counts within every range band. EviScope is a minimal test: the current frame's returns form a non-learned bird's-eye evidence field, which each query reads at its decoded centre to scale its radial sampling extent and to weight eight frames. It adds 5,450 parameters (0.009%) to RaCFormer, with no latency increase in our timing. Our module-free recipe exceeds the published TruckScenes mAP (25.9 over seven runs against 24.1), so we measure the module only against its own baseline at matched seeds. The evidence a query reads rises with its object's return count (Spearman 0.51) but captures zero-return objects only in part, and the learned radius tracks it ($ρ\approx-0.91$ in four runs). Over seven four-epoch pairs the module changes NDS by $+0.31\pm0.36$ and mAP by $-0.27\pm0.22$ (mean $\pm$ SE). Against a geometry-only conditioner with the same parameter count, trained at three matched seeds, it is higher by $+1.25\pm0.72$ NDS and $+0.87\pm0.21$ mAP, on mAP in every pair, and the control's mean NDS is near the baseline's. An exploratory far-range gain failed an internally pre-specified held-out test, and large inference-time substitution penalties also occur without query-local evidence: neither learned allocation nor inference-time dependence substitutes for matched end-to-end comparisons.

cs.CV

新论文 · v1arXiv:2611.25478

TAP: Tapping the Discriminative Potential of Frozen Representations for Class-Incremental Learning

Zihao Wang, Yu-Wei Zhan, Xiao-Qian Liu, Zhen-Duo Chen, Weitong Liu, Xin Luo, Xin-Shun Xu

原文摘要

Pre-trained models provide transferable representations for class-incremental learning (CIL). However, sequentially adapting these representations introduces optimization costs and may disrupt old-class feature structure. This motivates us to explore whether the discriminative potential of these representations can be further exploited by continually accumulating class information while keeping the encoders frozen. To this end, we propose TAP, a new and novel framework for tapping the discriminative potential of frozen representations in class-incremental learning. It accumulates visual and semantic class information in the frozen CLIP embedding space with only lightweight optimization. TAP comprises three modules: class-conditional visual distribution modeling, class-adaptive semantic calibration, and visual-semantic posterior fusion. The first module retains class means and within-class scatters, combining class-specific and shared covariances to stabilize distribution estimation. The second module learns class-specific weights over text templates from current-task images, adapting text representations to downstream classes. The third module combines the visual and semantic posteriors through a weighted product of experts for prediction over all seen classes without retaining images from previous tasks. Across nine datasets under two incremental protocols, TAP achieves state-of-the-art performance, attaining the highest average incremental and final-stage accuracies among the compared methods. The code is provided as an ancillary file.

cs.CV