Must-Read-LLM-Papers

KO | EN

gitlite — search

#javascript #python #hacktoberfest #ai #go #android #rust #deep-learning #docker #macos #java #awesome #cli #mcp

Must-Read-LLM-Papers

(★ 19)

No description available.

Must-Read-LLM-Papers Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Must-Read-LLM-Papers

## RL Optimization Algorithms

| Model    | Paper Title | arXiv Link |
|----------|-------------|------------|
| PPO      | Proximal Policy Optimization Algorithms | [1707.06347](https://arxiv.org/abs/1707.06347) |
| DPO      | Direct Preference Optimization: Your Language Model is Secretly a Reward Model | [2305.18290](https://arxiv.org/abs/2305.18290) |
| GRPO     | Generalized Reinforce Policy Optimization | [2501.12948](https://arxiv.org/abs/2501.12948) |
| DAPO     | Divergence-Augmented Preference Optimization | [2503.14476](https://arxiv.org/abs/2503.14476) |
| Dr.GRPO  | Distributionally Robust Generalized Reinforce Policy Optimization | [2503.20783](https://arxiv.org/abs/2503.20783) |
| StarPO   | Stabilized Reinforcement Alignment via Preference Optimization | [2504.20073](https://arxiv.org/abs/2504.20073) |
| ProRL    | Proximal Reinforcement Learning with Preference Feedback | [2505.24864](https://arxiv.org/abs/2505.24864) |
| GSPO     | Generalized Soft Preference Optimization | [2507.18071](https://arxiv.org/abs/2507.18071) |

## Deep Research

| Model        | Paper Title | arXiv Link |
|--------------|-------------|------------|
| R1-Searcher  | R1-Searcher: Training Reasoners by Searching with Language Models | [2503.05592](https://arxiv.org/abs/2503.05592) |
| Search-R1    | Search-R1: Scaling Reinforcement Learning with Tree Search for Reasoning | [2503.09516](https://arxiv.org/abs/2503.09516) |
| ReSearch     | ReSearch: Reinforcement Learning with Search for Reasoning | [2503.19470](https://arxiv.org/abs/2503.19470) |
| StepSearch   | StepSearch: Stepwise Search for Long-Horizon Reasoning | [2505.15107](https://arxiv.org/abs/2505.15107) |
| R1-Searcher++| R1-Searcher++: Scaling Search-Based Reinforcement Learning with Language Models | [2505.17005](https://arxiv.org/abs/2505.17005) |
| ZeroSearch   | ZeroSearch: Training Zero-Shot Reasoners via Search | [2505.04588](https://arxiv.org/abs/2505.04588) |
| WebDancer    | WebDancer: Reinforcement Learning with Web-Scale Exploration | [2505.22648](https://arxiv.org/abs/2505.22648) |
| WebSailor    | WebSailor: Autonomous Web Navigation with Reinforcement Learning | [2507.02592](https://arxiv.org/abs/2507.02592) |
| ASearcher    | ASearcher: Adaptive Search for Reasoning with Language Models | [2508.07976](https://arxiv.org/pdf/2508.07976) |
| SSRL         | SSRL: Self-Supervised Reinforcement Learning via Search | [2508.10874](https://arxiv.org/abs/2508.10874) |

## Foundation Models

| Model         | Paper Title                                                    | Link                                                                                |
| ------------- | -------------------------------------------------------------- | ----------------------------------------------------------------------------------- |
| GPT-3         | Language Models are Few-Shot Learners                          | [2005.14165](https://arxiv.org/abs/2005.14165)                                      |
| LLaMA         | LLaMA: Open and Efficient Foundation LLMs                      | [2302.13971](https://arxiv.org/abs/2302.13971)                                      |
| GPT-4         | GPT-4 Technical Report                                         | [2303.08774](https://arxiv.org/abs/2303.08774)                                      |
| LLaMA 2       | LLaMA 2: Open Foundation and Fine-Tuned Chat Models            | [2307.09288](https://arxiv.org/abs/2307.09288)                                      |
| Qwen          | Qwen Technical Report                                          | [2309.16609](https://arxiv.org/abs/2309.16609)                                      |
| Baichuan 2    | Baichuan 2: Open Large-Scale Language Models                   | [Tech Report PDF](https://cdn.baichuan-ai.com/paper/Baichuan2-technical-report.pdf) |
| Pythia        | Pythia: A Suite for Analyzing LLMs Across Training and Scaling | [GitHub](https://github.com/EleutherAI/pythia)                                      |
| Qwen2         | Qwen2 Technical Report                                         | [2407.10671](https://arxiv.org/abs/2407.10671)                                      |
| Qwen2.5       | Qwen2.5 Technical Report                                       | [2412.15115](https://arxiv.org/abs/2412.15115)                                      |
| DeepSeek-V3   | DeepSeek-V3 Technical Report                                   | [2412.19437](https://arxiv.org/pdf/2412.19437)                                      |
| DeepSeek-R1   | DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL | [2501.12948](https://arxiv.org/abs/2501.12948)                                      |
| Qwen2.5-VL    | Qwen2.5-VL Technical Report                                    | [2502.13923](https://arxiv.org/abs/2502.13923)                                      |
| Kimi K2   | Kimi K2: Open Agentic Intelligence                             | [2507.20534](https://arxiv.org/abs/2507.20534)                                      |
| Gemma 3       | Gemma 3 Technical Report                                       | [2503.19786](https://arxiv.org/pdf/2503.19786)                                      |
| Qwen3         | Qwen3 Technical Report                                         | [2505.09388](https://arxiv.org/abs/2505.09388)                                      |
| LongCat-Flash | LongCat-Flash Technical Report                                 | [2509.01322](https://arxiv.org/abs/2509.01322)                                      |

## LLM Interpretability

| Paper Title                                                                               | Author(s)             | Link                                                                                                                                                                         |
| ----------------------------------------------------------------------------------------- | --------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| INTERPRETABILITY IN THE WILD: A Circuit for Indirect Object Identification in GPT-2 Small | Anthropic             | [Distill](https://distill.pub/2020/circuits/)                                                                                                                                |
| Language Models Implement Simple Word2Vec-style Vector Arithmetic                         | Ellie Pavlick         | [Notion](https://www.notion.so/Language-Models-Implement-Simple-Word2Vec-style-Vector-Arithmetic-04b673b6c1e04a16b5b36d6524788305?pvs=21)                                    |
| Break It Down: Structural Compositionality in Neural Networks                             | Ellie Pavlick         | [Notion](https://www.notion.so/Break-It-Down-Evidence-for-Structural-Compositionality-in-Neural-Networks-2588066a47c3437a8accfbb022158926?pvs=21)                            |
| Does Circuit Analysis Interpretability Scale?                                             | Neel Nanda (DeepMind) | [Notion](https://www.notion.so/Does-Circuit-Analysis-Interpretability-Scale-Evidence-from-Multiple-Choice-Capabilities-in-Chinchil-97a4c9f3837644378819d71d71ec504e?pvs=21)  |
| Towards Automated Circuit Discovery                                                       | Neel Nanda            | [Notion](https://www.notion.so/Towards-Automated-Circuit-Discovery-for-Mechanistic-Interpretability-83327d70a2d54a63b90207582bcbe00e?pvs=21)                                 |
| TransformerLens (Toolkit)                                                                 | Neel Nanda            | [GitHub](https://github.com/neelnanda-io/TransformerLens)                                                                                                                    |
| Explaining Grokking via Circuit Efficiency                                                | DeepMind              | [2309.02390](https://arxiv.org/pdf/2309.02390.pdf)                                                                                                                           |

## LLM Scaling Law & Training Dynamics

| Paper Title                                 | Author(s) | Link                                                            |
| ------------------------------------------- | --------- | --------------------------------------------------------------- |
| Scaling Laws for Neural Language Models     | OpenAI    | [2001.08361](https://arxiv.org/abs/2001.08361)                  |
| Chinchilla: Training Compute-Optimal LLMs   | DeepMind  | [2203.15556](https://arxiv.org/abs/2203.15556)                  |
| Deep Double Descent                         | OpenAI    | [OpenAI Blog](https://openai.com/research/deep-double-descent)  |
| Emergent Abilities of LLMs                  | Jason Wei | [2206.07682](https://arxiv.org/abs/2206.07682)                  |
| Grokking: Generalization Beyond Overfitting | OpenAI    | [2201.02177](https://arxiv.org/abs/2201.02177)                  |
| Tensor Programs (series)                    | Greg Yang | [Tensor Programs](https://thegregyang.com/#tensorprograms)      |

## LLM Preference Alignment

| Paper Title                                        | Author(s)  | Link                                                                                                                                           |
| -------------------------------------------------- | ---------- | ---------------------------------------------------------------------------------------------------------------------------------------------- |
| Pretraining Language Models with Human Preferences | Sam Bowman | [Notion](https://www.notion.so/Pretraining-Language-Models-with-Human-Preferences-2d2851d8dfc04fa08ab2dfda4c3b45e6?pvs=21)                     |
| Alignment of Language Agents                       | DeepMind   | [Notion](https://www.notion.so/Alignment-of-Language-Agents-3eb77143e7f94aa5ada95254fb663f92?pvs=21)                                           |
| Evaluating the Ripple Effects of Knowledge Editing | Mor Geva   | [Notion](https://www.notion.so/Evaluating-the-Ripple-Effects-of-Knowledge-Editing-in-Language-Models-e323ab84eabd4a36a3bdc1f9d6348c51?pvs=21)  |
| Mass-Editing Memory in a Transformer               | David Bau  | [Notion](https://www.notion.so/Mass-Editing-Memory-in-a-Transformer-83e319401afa4941903a9a994218d856?pvs=21)                                   |
| RAIN: Self-Alignment without Fine-Tuning           | —          | [Notion](https://www.notion.so/RAIN-Your-Language-Models-Can-Align-Themselves-without-Finetuning-3d38d476991d4772b03eacde91c2c15f?pvs=21)      |

## Understanding LLMs

| Paper Title                                                                      | Author(s)        | Link                                                                                                                                                                         |
| -------------------------------------------------------------------------------- | ---------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Are Emergent Abilities in LLMs just In-Context Learning?                         | Google           | [2206.07682](https://arxiv.org/abs/2206.07682)                                                                                                                               |
| Language Models (Mostly) Know What They Know                                     | Anthropic        | [2304.05336](https://arxiv.org/abs/2304.05336)                                                                                                                               |
| Sparks of Artificial General Intelligence: Early Experiments with GPT-4          | Microsoft        | [2303.12712](https://arxiv.org/abs/2303.12712)                                                                                                                               |
| Unveiling Theory of Mind in LLMs: A Parallel to Neurons in the Human Brain       | —                | [Notion](https://www.notion.so/Unveiling-theory-of-mind-in-large-language-models-A-parallel-to-single-neurons-in-the-human-brain-bcdd472621d343f1903cb859a3e8d250?pvs=21)    |
| Delays, Detours, and Forks in the Road: Latent State Models of Training Dynamics | Kyunghyun Cho    | [Notion](https://www.notion.so/Delays-Detours-and-Forks-in-the-Road-Latent-State-Models-of-Training-Dynamics-231eef8a2fa7487583844b7824ef8992?pvs=21)                        |
| Generative Models as Complex Systems Science                                     | Luke Zettlemoyer | [Notion](https://www.notion.so/Generative-Models-as-a-Complex-Systems-Science-How-can-we-make-sense-of-large-language-model-behavi-9b2fe7aae69a4d438a166dcef6d51f1a?pvs=21)  |
| What Learning Algorithm is In-Context Learning?                                  | Tengyu Ma        | [2301.00234](https://arxiv.org/abs/2301.00234)                                                                                                                               |
| Why Can GPT Learn In-Context? Gradient Descent as Meta-Optimization              | Li Dong          | [Notion](https://www.notion.so/Why-Can-GPT-Learn-In-Context-Language-Models-Implicitly-Perform-Gradient-Descent-as-Meta-Optimizers-0d5b0e56043c4c82a2bd240fe0624a67?pvs=21)  |
| Measuring Inductive Biases of In-Context Learning                                | He He            | [Notion](https://www.notion.so/Measuring-Inductive-Biases-of-In-Context-Learning-with-Underspecified-Demonstrations-e4a4f05eec924563a6afc2b4d74405b1?pvs=21)                 |
| Faith and Fate: Limits of Transformers on Compositionality                       | Yejin Choi       | [Notion](https://www.notion.so/Faith-and-Fate-Limits-of-Transformers-on-Compositionality-12c4d8b6b37540aa87ebeb8140e0f474?pvs=21)                                            |
| Intrinsic Dimensionality Explains Effectiveness of LM Fine-Tuning                | Luke Zettlemoyer | [Notion](https://www.notion.so/Intrinsic-Dimensionality-Explains-the-Effectiveness-of-Language-Model-Fine-Tuning-9d72262a131d46eb93bf512e34e2232e?pvs=21)                    |

## Reward Models & Benchmarks

| Method / Benchmark | Full Title                                                               | Link                                                              |
| ------------------ | ------------------------------------------------------------------------ | ----------------------------------------------------------------- |
| Checklists         | Checklists Are Better Than Reward Models For Aligning Language Models    | [2507.18624](https://arxiv.org/pdf/2507.18624)                    |
| RewardAnything     | RewardAnything: Generalizable Principle-Following Reward Models          | [2506.03637](https://arxiv.org/abs/2506.03637)                    |
| RubricRM           | Rubrics as Rewards: Reinforcement Learning Beyond Verifiable Domains     | [2507.17746](https://arxiv.org/abs/2507.17746)                    |
| RM-R1              | RM-R1: Reward Modeling as Reasoning                                      | [2505.02387](https://arxiv.org/abs/2505.02387)                    |
| GenRM              | Generative Reward Models                                                 | [2410.12832](https://arxiv.org/abs/2410.12832)                    |
| RM-Bench           | RM-Bench: Benchmarking Reward Models for Language Model Alignment        | [2406.11695](https://arxiv.org/abs/2406.11695)                    |
| RewardBench        | RewardBench: A Benchmark for Evaluating Reward Models in LLM Alignment   | [RewardBench](https://huggingface.co/spaces/allenai/reward-bench) |
| Skywork-Reward-V2  | Skywork-Reward-V2: Scaling Preference Data Curation via Human-AI Synergy | [2507.01352](https://arxiv.org/abs/2507.01352)                    |
| Deepseek-GRM       | Inference-Time Scaling for Generalist Reward Modeling                    | [2504.02495](https://arxiv.org/abs/2504.02495)                    |
| Agent-as-a-Judge   | Agent-as-a-Judge: Evaluate Agents with Agents                            | [2410.10934](https://arxiv.org/abs/2410.10934)                    |

## Agent Benchmark

| Benchmark  | Full Title                                                | Link                                                        |
| ---------- | --------------------------------------------------------- | ----------------------------------------------------------- |
| Tau-Bench  | Tau-Bench: A Benchmark for Tool-Augmented Language Models | [Tau-Bench](https://tau-bench.github.io/)                   |
| Tau2-Bench | Tau2-Bench: Benchmarking General-Purpose Language Agents  | [Tau2-Bench](https://tau-bench.github.io/tau2/)             |
| MCP-Bench  | MCP-Bench: Benchmarking Tool-Using LLM Agents with Complex Real-World Tasks via MCP Servers     | [2508.20453](https://arxiv.org/pdf/2508.20453v1) |

Must-Read-LLM-Papers

Must-Read-LLM-Papers Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Must-Read-LLM-Papers

Must-Read-LLM-Papers Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README