- 博客(4672)
- 资源 (1)
- 收藏
- 关注
RSS订阅原创 Beyond Refusal Patterns 论文总结与翻译
# Beyond Refusal Patterns 论文总结与翻译> 论文信息：arXiv:2610.07023 | Beyond Refusal Patterns: Safe‑Role Internalization for Robus
2026-10-31 12:00:00
1
原创 Distributionally Robust Mixture‑of‑Experts Training（DRMoET）论文总结与翻译
# Distributionally Robust Mixture‑of‑Experts Training（DRMoET）论文总结与翻译> 论文信息：arXiv:2610.07207，NeurIPS 2026，作者：Xin Teng, M
2026-10-31 08:00:00
78
原创 CoDe‑LoRA 论文阅读总结
# CoDe‑LoRA 论文阅读总结> 论文信息：CoDe‑LoRA: Mitigating the Orthogonality Dilemma in Continual Learning of LLMs via Knowledge Co
2026-10-31 07:00:00
84
原创 Dynamic Positional Attention Modulation 论文阅读笔记
# Dynamic Positional Attention Modulation 论文阅读笔记> 论文信息：arXiv:2610.07848，KDD 2026，标题：*Dynamic Positional Attention Modul
2026-10-30 07:00:00
69
原创 Illusory Pattern Perception Drives Spurious Inference in Large Language Models 论文总结与译文
# Illusory Pattern Perception Drives Spurious Inference in Large Language Models 论文总结与译文> 论文信息：arXiv:2610.07791，NeurIPS
2026-10-29 09:00:00
7
原创 Understanding and Enhancing Backdoor Persistency in LLM Agent Post‑Training 论文总结与翻译
# Understanding and Enhancing Backdoor Persistency in LLM Agent Post‑Training 论文总结与翻译> 论文arXiv：https://arxiv.org/abs/26
2026-10-29 08:00:00
100
原创 MILO: Efficient Many‑shot In‑Context Learning with Block‑wise Low‑rank Compression
> 说明：arXiv摘要页面仅公开**Abstract**，未直接展示Introduction、Conclusion完整原文；PDF需要读取完整论文。当前可获取完整的只有Abstract，因此我完成：论文整体总结、创新点、Abstract中
2026-10-29 07:00:00
128
原创 Rationale‑Guided Policy Optimization（RGPO）论文总结与翻译
# Rationale‑Guided Policy Optimization（RGPO）论文总结与翻译> 论文信息：arXiv:2610.07342｜NeurIPS 2026｜[论文仓库](https://github.com/VietH
2026-10-28 13:00:00
7
原创 Component and Dimension Sparsity in Transformer Refusal Mechanisms 论文总结
# Component and Dimension Sparsity in Transformer Refusal Mechanisms 论文总结> 论文信息：arXiv:2610.06903，接收于COLM 2026，作者：Vincen
2026-10-28 12:00:00
8
原创 CodeGraph: Open‑Taxonomy Knowledge Graph for Source Code with Wikidata Grounding
# CodeGraph: Open‑Taxonomy Knowledge Graph for Source Code with Wikidata Grounding> 论文arXiv链接：[https://arxiv.org/abs/26
2026-10-28 09:00:00
89
原创 LogicTrack: Auditing Reasoning Trajectories of Large Language Models with Formal Logic Solvers 论文总结与
# LogicTrack: Auditing Reasoning Trajectories of Large Language Models with Formal Logic Solvers 论文总结与翻译> 论文arXiv链接：htt
2026-10-28 08:00:00
101
原创 Accelerating Dense LLMs via L0‑regularized Mixture‑of‑Experts 论文总结与翻译
# Accelerating Dense LLMs via L0‑regularized Mixture‑of‑Experts 论文总结与翻译> 论文arXiv链接：https://arxiv.org/abs/2609.21672##
2026-10-28 07:00:00
8
原创 Abstention and Noise Filtering: Two Missing Primitives of Softmax Attention 论文阅读总结
# Abstention and Noise Filtering: Two Missing Primitives of Softmax Attention 论文阅读总结> 论文arXiv链接：[https://arxiv.org/abs/
2026-10-27 13:00:00
9
原创 Block‑Sparse Attention with Semantic‑Geometric Decoupled Routing（arXiv:2609.22884）论文总结与翻译
# Block‑Sparse Attention with Semantic‑Geometric Decoupled Routing（arXiv:2609.22884）论文总结与翻译> 论文链接：[https://arxiv.org/ab
2026-10-27 12:00:00
7
原创 GVPO++: Group Variance Policy Optimization for LLM Post‑Training and On‑Policy Distillation
# GVPO++: Group Variance Policy Optimization for LLM Post‑Training and On‑Policy Distillation> 论文链接：https://arxiv.org/a
2026-10-27 10:00:00
298
原创 RBS‑Attention 论文阅读总结
# RBS‑Attention 论文阅读总结> 论文信息：arXiv:2609.20971，RBS‑Attention: Radius‑Bounded Sparse Prefill for Long‑Context Large Langu
2026-10-27 09:00:00
94
原创 Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning（arXiv:2609.03430）
# Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning（arXiv:2609.03430）## 一、文章主要内容总结### 研究背景大模型在长链
2026-10-27 08:00:00
135
原创 SWE‑Gate 论文阅读总结
# SWE‑Gate 论文阅读总结> 论文信息：arXiv:2609.04167，SWE‑Gate: Passing Functional Tests Is Not Enough for Software Engineering Agen
2026-10-27 07:00:00
10
原创 What Matters for Aggressive Decoding‑Time KV Eviction? Temporal Aggregation and Ranking Preservation
# What Matters for Aggressive Decoding‑Time KV Eviction? Temporal Aggregation and Ranking Preservation> arXiv:2609.0351
2026-10-26 15:00:00
8
原创 GrowPage: On‑Demand KV Budgeting for Efficient LLM Reasoning Serving（arXiv:2609.03494）
# GrowPage: On‑Demand KV Budgeting for Efficient LLM Reasoning Serving（arXiv:2609.03494）> 论文链接：[https://arxiv.org/abs/2
2026-10-26 14:00:00
8
原创 Large Language Models in Resolving Contextual Knowledge Conflicts（arXiv:2609.03148）
# Large Language Models in Resolving Contextual Knowledge Conflicts（arXiv:2609.03148）> 会议：EMNLP 2026> 论文链接：https://arx
2026-10-26 13:00:00
6
原创 RuleMem: Active Rule Memory for Long‑Term Conversational Agents 论文总结与翻译
# RuleMem: Active Rule Memory for Long‑Term Conversational Agents 论文总结与翻译> 论文arXiv链接：https://arxiv.org/abs/2609.03915
2026-10-26 12:00:00
7
原创 Representational alignment yields generalizable safety in language models
> 说明：arXiv网页摘要接口仅公开**Abstract**，**Introduction、Conclusion**正文内容不在返回的元数据中，无法直接获取原文，因此仅完成摘要原文+翻译；文章主要内容、创新点基于摘要信息整理。# Re
2026-10-26 11:00:00
109
原创 It‘s the Problem, Not the Path: Budget and Difficulty Confounds in LLM Reasoning Trajectories
# It's the Problem, Not the Path: Budget and Difficulty Confounds in LLM Reasoning Trajectories> 论文arXiv地址：https://arxi
2026-10-26 10:00:00
140
原创 论文总结：Doesn‘t Stop Reasoning: Analysis of Spurious CoT Termination
# 论文总结：Doesn't Stop Reasoning: Analysis of Spurious CoT Termination> 论文arXiv：https://arxiv.org/abs/2609.03633> 会议：EMNL
2026-10-26 09:00:00
211
原创 GeoEdit: Geometric Knowledge Editing for Large Language Models
### 摘要定期更新对于保持大语言模型（LLMs）中的知识时效性至关重要。因此，人们开发了各种模型编辑方法来更新 LLMs 中的特定知识。然而，基于训练的方法往往难以在有效融入新知识的同时保留不相关的通用知识。为应对这一挑战，我们提出了
2026-10-26 08:00:00
5
原创 Survey of Large Multimodal Reasoning Models (LMRMs): From Modular to Native (arXiv:2505.04921)
### 文章总结 本文是哈尔滨工业大学（深圳）团队发表的关于大型多模态推理模型（LMRMs）的综述论文，系统梳理了多模态推理模型的发展脉络、技术挑战及未来方向。核心内容如下：#### **主要内容** 1. **多模态推理模型
2026-10-26 07:00:00
8
原创 MemoryLACE论文总结与翻译
# MemoryLACE论文总结与翻译> 论文信息：arXiv:2609.03201，MemoryLACE: Memory Lifecycle‑Aware Consolidation and Evidence Retrieval##
2026-10-25 16:00:00
9
原创 FlowBalance 论文阅读笔记
# FlowBalance 论文阅读笔记> 论文信息：FlowBalance: Verifier‑Grounded Self‑Improvement from On‑Policy Reasoning Experience，arXiv:26
2026-10-25 15:00:00
7
原创 The Geometr\(\boldsymbol{y}\) of Ignorance: LLMs Know \(\bolds\(\boldsymbol{y}\)mbol{\(\boldsymbol{W
# The Geometr\(\boldsymbol{y}\) of Ignorance: LLMs Know \(\bolds\(\boldsymbol{y}\)mbol{\(\boldsymbol{W}\)}\)hen to Tempe
2026-10-25 14:00:00
9
原创 ALRA: Adaptive Local Relational Alignment 论文总结与翻译
# ALRA: Adaptive Local Relational Alignment 论文总结与翻译> 论文信息：arXiv:2609.03355，基于Logit的自回归语言模型预训练蒸馏自适应局部关系对齐方法## 一、文章主要内容
2026-10-25 13:00:00
12
原创 Dream‑RSI: Recursive Self‑Improvement through Evolving Worlds 论文总结与翻译
# Dream‑RSI: Recursive Self‑Improvement through Evolving Worlds 论文总结与翻译> 论文信息：arXiv:2609.14858（v2，2026‑10‑06）> 作者：Tong
2026-10-25 12:00:00
13
原创 Meta-Reasoner: Dynamic Guidance for Optimized Inference-time Reasoning in Large Language Models
### 摘要大语言模型（LLMs）越来越依赖冗长的推理链来解决复杂任务。然而，这种试错方法往往会导致高昂的计算成本和错误传播，早期的错误可能会使后续步骤偏离正轨。为解决这些问题，我们引入了 Meta-Reasoner 框架，它通过让大语
2026-10-25 11:00:00
111
原创 Collaborative Stance Detection via Small-Large Language Model Consistency Verification
### 摘要社交媒体上的立场检测旨在识别推文中对特定目标表达的态度。当前研究由于大语言模型（LLMs）具有显著的性能提升，因而更倾向于使用它们，而非小语言模型（SLMs）。然而，对于需要大量数据分析的现实社交媒体监测系统而言，不计成本地
2026-10-25 10:00:00
125
原创 1-Shot RLVR: Reinforcing LLM Math Reasoning with a Single Example (arXiv:2504.20571)
### 主要内容1. **研究背景**：强化学习与可验证奖励（RLVR）在提升大语言模型（LLMs）推理能力方面取得显著进展，但RLVR数据方面的研究相对不足，如训练数据的数量、质量和有效性等问题有待探索。2. **研究方法**
2026-10-25 09:00:00
9
原创 Interactive Generative Video (IGV): Survey for Gaming, Embodied AI and Autonomous Driving (arXiv:250
### 主要内容1. **交互式生成视频（IGV）概述**：IGV是结合生成能力与交互功能的技术，能生成多样高质量视频内容，并通过控制信号和反馈实现用户参与。文章介绍了视频生成的基础技术，如VAE、GAN、扩散模型、自回归模型及混合模型等
2026-10-25 08:00:00
12
原创 APRMCTS: Improving LLM-based Automated Program Repair with Iterative Tree Search
自动程序修复（APR）旨在无需人工干预即可修复软件漏洞，在软件开发和维护中发挥着关键作用。近年来，随着大语言模型（LLMs）的发展，大量基于LLM的APR技术被提出，并取得了显著性能。然而，现有基于LLM的APR技术通常采用试错策略，存在两大缺陷：（1）由于局部探索，补丁有效性存在固有局限；（2）由于冗余探索，搜索效率低下。本文提出了APRMCTS，该方法通过迭代树搜索改进基于LLM的APR。
2026-10-25 07:00:00
12
原创 Legibility is Not Interpretability: Comparing Judged and Actual Importance in Chain‑Of‑Thought Reaso
# Legibility is Not Interpretability: Comparing Judged and Actual Importance in Chain‑Of‑Thought Reasoning> 论文信息：arXiv:
2026-10-24 16:00:00
8
原创 SoL‑Pi: Recursively Scaling Auto‑Research Loops for Efficient Agent Harness 论文总结与翻译
# SoL‑Pi: Recursively Scaling Auto‑Research Loops for Efficient Agent Harness 论文总结与翻译> 论文arXiv：https://arxiv.org/abs/26
2026-10-24 15:00:00
12
原创 论文阅读笔记：Grounded Skill Synthesis from Code at Scale for Agentic Intelligence
# 论文阅读笔记：Grounded Skill Synthesis from Code at Scale for Agentic Intelligence> arxiv：https://arxiv.org/abs/2609.05571>
2026-10-24 14:00:00
14
空空如也
TA创建的收藏夹 TA关注的收藏夹
TA关注的人