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

+1680新增论文
+1333其中 AI 相关
+187已有论文修订
+59新增 AI 导读

arXiv 最近完成扫描:2026-11-18 04:28 · 全站最近入库:2026-11-17 23:35各信源进度 →

统计口径与数据新鲜度

统计 2026-11-17 09:16 — 2026-11-18 09:16(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:2026-11-18 04:28最近发现新增版本:2026-11-17 23:35库内论文最新更新时间:2026-11-17 02:59

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。

25,786 已收录论文21,491 AI 相关7,725 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 3 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-18 04:27(北京时间)
版本更新 · v3arXiv:2609.39082

Shared Phase and Retention Control for Efficient Adaptive Spectral Recurrence

Wentao Wang, Hengyu Zhong, Yunhan Jiang, Jialiang An, Meng Lu

原文摘要

Fixed-state spectral recurrences offer constant-memory inference and strong long-range retention through heterogeneous complex modes, but their static transitions cannot selectively filter distractors or adapt to evolving dynamics. Existing input-dependent recurrences introduce selectivity, yet real-valued gates lack adaptive rotation, while fine-grained complex transitions incur substantial projection and data-movement overhead and can expose persistent modes to cumulative drift. To address these limitations, we introduce Shared Phase and Retention Control for Efficient Adaptive Spectral Recurrence (SPARC), which modulates a layer's heterogeneous complex memory through two input-dependent scalars. Each mode retains its learned baseline timescale, frequency, and write response, while a retention-coupled clock damps phase corrections in persistent modes. The same shared signals regulate incoming content through modal gates and retention-dependent normalization. The resulting diagonal affine recurrence supports parallel associative scans for full-sequence backpropagation through time (BPTT) and exact structured Real-Time Recurrent Learning (RTRL), with sensitivity storage and updates linear in the recurrent layer's parameter count. We evaluate SPARC on partially observable continuous control, POPGym memory benchmarks, sequence classification, and language modeling. SPARC places among the top two means across all ten reinforcement-learning tasks. Compared with RG-LRU, it improves permuted-MNIST and sequential-CIFAR-10 accuracy by $10.13$ and $10.66$ percentage points, respectively, reduces held-out language-model perplexity by $2.00\%$, and accelerates mixer forward--backward computation by up to $1.69\times$ at width $D=2048$. Code is available at https://github.com/Botwwt/sparc.

cs.LG

新论文 · v1arXiv:2611.02432

Diagnosing Model--Item Interactions in Language Model Benchmarks with LSIRM

Jinwen Luo, Han Jiang, Ziang Xiao, Susu Zhang

原文摘要

Language models with similar benchmark accuracy can succeed on different items. Overall scores do not show where their performance differs or which item properties might explain those differences. We demonstrate how the latent space item response model (LSIRM) can help evaluators investigate these patterns and their consequences for model comparison. We analyze 1,348 language models on 5,761 BBH items and compare ways of presenting and scoring answers across BBH, GPQA, MMLU-Pro, and MuSR. We examine LSIRM maps, which summarize differences in performance across models and items, alongside item properties and records of correct and incorrect responses. In BBH, nearby items on these maps often share the same correct answer, such as "Yes" or "No." Across the four benchmarks, this pattern is stronger when answers are scored using option labels, such as A, B, C, and D, than when using the text of the options. We then estimate each model's tendency to perform better or worse on items with particular correct answers and quantify how this tendency contributes to its score. This separates an estimated contribution associated with correct answers from the remaining differences in item performance. Two models' total numbers of correct answers differed by only 14 out of 5,761 items. Excluding the estimated answer-category contribution from their scores moved their ranks hundreds of places in opposite directions, while keeping all other estimates unchanged. The approach helps evaluators identify differences hidden by overall accuracy and understand how the choice of what to include in a score affects model rankings.

cs.CL · cs.LG · stat.ME

版本更新 · v2arXiv:2607.28862

TextCloak: Safeguarding Data from Unauthorized LLM Fine-Tuning via RL-Driven Unlearnable Examples

Chengshuai Zhao, Pingchuan Ma, Dawei Li, Bohan Jiang, Zhiyuan Yu, Zhen Tan, Huan Liu

原文摘要

The development of Large Language Models (LLMs) has led to significant advances across a wide range of language tasks, while simultaneously raising growing concerns about unauthorized online data exploitation and privacy leakage. Unlearnable examples (UEs) emerge as a promising defense by injecting crafted perturbations into data such that models trained on them exhibit degraded utility. However, existing methods for text protection are primarily designed for classification tasks and discriminative language models, which limits their broad applicability (e.g., instruction following or reasoning) in LLMs. In this work, we propose **TextCloak**, an RL-driven data synthesis framework that safeguards textual data against unauthorized LLM fine-tuning. TextCloak employs a *UE generative policy* that transforms clean text into unlearnable examples while preserving semantic fidelity and linguistic naturalness. To optimize the policy, we introduce **GRPO-UE**, which simulates the attacker's LLM fine-tuning and rewards downstream degradation via group-relative policy optimization. This overall formulates a bi-level optimization loop, allowing the generator to discover effective protective patterns. Comprehensive experiments on six publicly available datasets and nine state-of-the-art LLMs demonstrate that TextCloak consistently impairs unauthorized LLM fine-tuning while maintaining text utility for legitimate use. Further analyses confirm its transferability and robustness across model architectures, training configurations, and adaptive attacks, highlighting its broad applicability as a practical defense for online content.

cs.CL · cs.AI · cs.CR · cs.LG