跳转至

论文笔记

共 42 篇（21 篇已精读，21 篇暂无笔记）。点击标题进入论文页；行尾「笔记」「译文」直达对应内容（缺失则置灰）。

Verification

verification · 11 篇

★★★★★2607.05391v2
LLM-as-a-Verifier: A General-Purpose Verification Framework
Stanford University；UC Berkeley；NVIDIA Research
Trajectory Scoring
笔记译文
2026-07-28
★★★★★2510.13744v1
Hard2Verify: A Step-Level Verification Benchmark for Open-Ended Frontier Math
ACL 2026（Long） · Salesforce AI Research
EvaluationMath
笔记译文
2026-07-03
★★★★★2601.15808v2
Inference-Time Scaling of Verification: Self-Evolving Deep Research Agents via Test-Time Rubric-Guided Verification
Findings of ACL 2026 · The Chinese University of Hong Kong；Tencent AI Lab；Singapore Management University；Renmin University of China
Deep ResearchRubric-guided
笔记译文
2026-07-03
★★★★★2509.17995
Variation in Verification: Understanding Verification Dynamics in Large Language Models
ICLR 2026 · Salesforce AI Research；Dartmouth College；University of Illinois Urbana-Champaign
Evaluation
笔记译文
2026-07-03
★★★★★2509.23152
Critique to Verify: Accurate and Honest Test-Time Scaling with RL-Trained Verifiers
The Hong Kong University of Science and Technology (Guangzhou)；The Hong Kong University of Science and Technology；ETH AI Center, ETH Zurich；University of California, Merced；Sun Yat-sen University；MBZUAI
MathCritique-based
笔记译文
2026-07-03
★★★★★2602.04254
Scaling Agentic Verifier for Competitive Coding
Renmin University of China；Qwen Team, Alibaba Group
CodingCounterexample Search
笔记译文
2026-07-03
★★★★★2110.14168
Training Verifiers to Solve Math Word Problems
OpenAI
MathOutcome Verifier
笔记译文
2026-07-02
★★★★★2211.14275
Solving math word problems with process- and outcome-based feedback
DeepMind
MathProcess vs Outcome
笔记译文
2026-07-02
★★★★★2305.20050
Let's Verify Step by Step
ICLR 2024 · OpenAI
MathProcess Reward Model
笔记译文
2026-07-02
★★★★★2311.09724
OVM, Outcome-supervised Value Models for Planning in Mathematical Reasoning
NAACL Findings · The Chinese University of Hong Kong, Shenzhen；Shenzhen Research Institute of Big Data
MathValue Model
笔记译文
2026-07-02
★★★★★2508.19903
Logical Reasoning with Outcome Reward Models for Test-Time Scaling
EMNLP 2025 · Monash University；VinUniversity
LogicOutcome Reward Model
笔记译文
2026-07-02

创意写作 / 叙事

narrative · 4 篇

★★★★★2601.09609
DPWriter：面向创意写作的多样化规划分支强化学习
中国人民大学；快手科技（一作在快手实习期间完成）
RL-trainedPlanning
笔记译文
2026-07-27
★★★★★2504.11900
Finding Flawed Fictions：通过情节漏洞检测评估语言模型的复杂推理
University of Washington，Paul G. Allen School of Computer Science & Engineering
Evaluation
笔记译文
2026-07-28
★★★★★2603.05890
Lost in Stories：LLM 长篇故事生成中的一致性错误
Microsoft Beijing；Singapore University of Technology and Design
Evaluation
笔记译文
2026-07-28
★★★★★2512.02240
Lightweight Latent Reasoning for Narrative Tasks
University of Edinburgh；CIFAR Fellow
Latent Reasoning
笔记译文
2026-07-21

评测与 Benchmark

evaluation · 1 篇

★★★★★2411.04872v7
FrontierMath: A Benchmark for Evaluating Advanced Mathematical Reasoning in AI
Epoch AI；贡献数学家来自 King's College London、MIT、University of Siegen、ICMC/USP、UC Berkeley、Cornell、Harvard、University of Leicester、University of Bristol、Rutgers、ETH Zurich 等机构
Math
笔记译文
2026-07-16

Agent RL

agent_rl · 1 篇

★★★★★2607.07508
SAO 精读：让长时程 Agent 的异步 RL 真正训得动
Tsinghua University
Asynchronous RLSingle Rollout
笔记译文
2026-07-13

RL / Reward Learning

general_rl · 2 篇

★★★★★2410.18451
Skywork-Reward: Bag of Tricks for Reward Modeling in LLMs
Skywork AI, Kunlun Inc.
Preference RM
笔记译文
2026-07-06
★★★★★2502.06781
Exploring the Limit of Outcome Reward for Learning Mathematical Reasoning
Shanghai AI Laboratory；Shanghai Jiao Tong University；MMLab, The Chinese University of Hong Kong；HKGAI under InnoHK
VerificationMath
笔记译文
2026-07-02

视频生成

video_generation · 1 篇

★★★★★2606.17800
MaineCoon 精简解析
Catnip AI Team
Real-time GenerationAudio-visual
笔记译文
2026-07-03

幻灯片生成

slides · 1 篇

★★★★★2606.17162
MemSlides 精简解析
Beijing University of Posts and Telecommunications；Tsinghua University；Shanghai Jiao Tong University
PersonalizationHierarchical Memory
笔记译文
2026-07-02

待精读

pending · 21 篇

★★★★★2502.10391
MM-RLHF：多模态 LLM 对齐的下一步
笔记译文
★★★★★2503.05244
WritingBench：面向生成式写作的综合基准
笔记译文
★★★★★2504.03622
与结构对齐：利用结构信息对齐大语言模型
笔记译文
★★★★★2506.00103
2506.00103_writing_zero
笔记译文
★★★★★2506.05760
2506.05760_writing_rl
笔记译文
★★★★★2508.18642
2508.18642_rlmr
笔记译文
★★★★★2508.21476
2508.21476_ignite_creative_slm
笔记译文
★★★★★2509.02534
联合强化语言模型生成中的多样性与质量
笔记译文
★★★★★2511.10643
Black-Box On-Policy Distillation of Large Language Models
笔记译文
★★★★★2511.22570
DeepSeekMath-V2：迈向可自验证的数学推理
笔记译文
★★★★★2601.07149
2601.07149_rewarding_creativity
笔记译文
★★★★★2601.09858
OUTLINEFORGE：面向科学写作、具有显式状态的层次化强化学习
笔记译文
★★★★★2601.17226
2601.17226_retell_reward_repeat
笔记译文
★★★★★2602.01511
2602.01511_rubric_arm
笔记译文
★★★★★2603.15061
2603.15061_writer_r1
笔记译文
★★★★★2604.03004
2604.03004_r2_write
笔记译文
★★★★★2604.05517
2604.05517_unicreative
笔记译文
★★★★★2604.19071
HoWToBench：用“写作树”整体评估大语言模型的人类水平写作能力
笔记译文
★★★★★2604.27453
2604.27453_coarse_to_fine
笔记译文
★★★★★2606.26300
验证地平线：编码 Agent 奖励没有银弹
笔记译文
★★★★★2606.29354
When LLMs Develop Languages：用于高效多智能体推理的符号通信
笔记译文