論文標題：Reasoning with Language Model is Planning with World Model論文鏈接：https://aclantholog...
發(fā)簡信
IP屬地：北京
-
從 CoT 到 RAP：用世界模型增強大模型推理 -
Memory-R1：用強化學習讓大模型智能體學會管理長期記憶論文標題：Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories vi...
-
Reflexion：讓語言智能體通過語言反饋自我強化論文標題：Reflexion: Language Agents with Verbal Reinforcement Learning論文鏈接：https://arxiv.or...
-
ToolRL：工具調用不是格式模仿，而是獎勵學習論文標題：ToolRL: Reward is All Tool Learning Needs論文鏈接：https://arxiv.org/abs/2504.13958[htt...
-
ACL 2025 - 基于片段監(jiān)督偏好優(yōu)化的字幕翻譯時延對齊論文標題：Fine-grained Video Dubbing Duration Alignment with Segment Supervised Preference O...
-
直接偏好優(yōu)化技術DPO基礎理論及推導
論文標題：Direct Preference Optimization: Your Language Model is Secretly a Reward Model論文鏈接...
-
自適應視圖增強的謠言檢測圖對比學習方法論文標題：Propagation Tree Is Not Deep: Adaptive Graph Contrastive Learning Approach for Rum...
-
生成式大模型的RLHF技術（一）：基礎一、概述 大語言模型（LLMs）在預訓練的過程中通常會捕捉數(shù)據(jù)的特征，而這些訓練數(shù)據(jù)通常既包含高質量的也包含低質量的，因此模型有時會產生不被期望的行為，如編造事實，生成有偏見...
-
LoRA：大模型下游任務的低秩適應論文標題：LoRA: Low-Rank Adaptation of Large Language Models論文鏈接：https://arxiv.org/abs/2106....
-
Megatron-LM：Transformer模型專用分布式張量模型并行方法論文標題：Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallel...
-
思維樹：大模型的復雜推理技術論文標題：Tree of Thoughts: Deliberate Problem Solving with Large Language Models論文鏈接：https:...
-
LIMA：小規(guī)模監(jiān)督數(shù)據(jù)指令微調論文標題：LIMA: Less Is More for Alignment論文鏈接：https://arxiv.org/abs/2305.11206[https://arxi...
-
語言模型的自洽性思維鏈推理技術論文標題：Self-Consistency Improves Chain of Thought Reasoning in Language Models論文鏈接：https:...
-
GPipe：微批量流水線并行論文標題：GPipe: Easy Scaling with Micro-Batch Pipeline Parallelism論文鏈接：https://arxiv.org/ab...
-
InstructGPT：語言模型的人類反饋指令對齊
論文標題：Training language models to follow instructions with human feedback論文鏈接：https://ar...
-
高效底座模型LLaMA論文標題：LLaMA: Open and Efficient Foundation Language Models論文鏈接：https://arxiv.org/abs/230...
-
TokenGT：Transformer是強大的圖學習器論文標題：Pure Transformers are Powerful Graph Learners論文鏈接：https://arxiv.org/abs/2207.02505...