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

+1959新增论文
+1550其中 AI 相关
+212已有论文修订
+59新增 AI 导读

arXiv 最近完成扫描:2026-11-17 23:20 · 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 03:17 — 2026-11-18 03:17(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:2026-11-17 23:20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 2 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-17 23:19(北京时间)
新论文 · v1arXiv:2611.20288

Protecting Safety during Reasoning Training with Low-Rank Fisher Regularisation

Neel Rajani, Ivan Titov

原文摘要

Large language models are trained to have safety guardrails, but these can be compromised while fine-tuning for new capabilities, such as reasoning. One way to mitigate this degradation is to modify the fine-tuning objective to preserve safety behaviour, but doing so effectively and efficiently remains challenging. We revisit Elastic Weight Consolidation (EWC), which penalises parameter updates based on sensitivity estimates from the diagonal of the Fisher information matrix (FIM). While EWC provides some protection, the diagonal approximation ignores interactions between parameters. We address this limitation by exploiting recent findings that safety-relevant behaviour is located in a low-rank subspace within models. Our method, Subspace-Projected Elastic Weight Consolidation (SP-EWC), projects gradients onto this subspace to capture non-diagonal curvature efficiently. We investigate three different ways of identifying the safety subspace: through safety fine-tuning updates (LoRA), safety-related activations (ActSVD), and contrastive representations of safety and utility data (cPCA). We find that this choice substantially affects performance, with SP-EWC using cPCA comparing favourably against existing methods by better maintaining safety guardrails without introducing excessive over-refusal.

cs.LG · cs.AI

版本更新 · v2arXiv:2605.29075

知识卸载:将大语言模型分解为稀疏主干与记忆模块

Knowledge Offloading: Decomposing LLMs into Sparse Backbones and Memory Modules

Karim Galliamov, Rochelle Choenni, Ivan Titov

AI 速读 基于摘要

作者提出 KOFF 框架,尝试把预训练大模型拆成共享稀疏主干与领域专属外部记忆,并报告可在较大程度保留困惑度与 MMLU 表现的同时移出部分容量。

cs.LG

查看原文摘要

LLMs encode both general capabilities and domain-specific knowledge in a single set of parameters. We ask whether this capacity can be reorganized: keeping broadly useful computation in a shared backbone, while moving specialized knowledge into external memory modules. We propose \emph{knowledge offloading} (KOFF), a framework for decomposing a pretrained LLM into a sparse shared backbone and domain-specific memories. Starting from a frozen base model, we either jointly or sequentially learn a structured pruning mask and lightweight recovery modules, implemented as LoRA adapters and learned key-value caches. Across Llama and Qwen models from 3B to 14B, we find that 12\% to 20\% of capacity can be moved out of the shared backbone while largely preserving perplexity and MMLU performance (up to 90\% of the full model). We show that LoRA and learned KV memories are complementary, and analyses suggest that the offloaded knowledge is specialized: mismatched memories do not yield the same gains, and language-specific neurons are preferentially removed while language-neutral neurons largely remain in the backbone. These results suggest that the knowledge of a pretrained LLM can be reallocated between a shared core and swappable external memories.