Awesome Agents Papers Collection

A comprehensive collection of papers and presentation slides on LLM agents, reasoning, and AI systems.

Sources:

• arvindcr4/awesome-agents
• redhat-et/agentic-reasoning-reinforcement-fine-tuning

Quick Stats

• Papers: 88 PDFs (organized in 12 folders)
• Slides: 93 presentation decks (~504 MB)
• Topics: 15 categories
• Audio Overviews: See NOTEBOOKLM_LINKS.md for AI-generated podcast summaries
• Resources: See DEEP_RL_RESOURCES.md for comprehensive RL learning materials

Folder Structure

papers/
├── agent-frameworks/    # 10 papers - ReAct, AutoGen, DSPy, etc.
├── benchmarks/          #  6 papers - SWE-bench, WorkArena, evals
├── computer-use/        #  5 papers - OSWorld, DigiRL, SWE-agent
├── memory-rag/          #  3 papers - HippoRAG, retrieval systems
├── multi-agent/         #  2 papers - AgentNet, MasRouter
├── planning/            #  5 papers - Tree search, optimization
├── reasoning/           #  9 papers - Chain-of-thought, reasoning
├── rl-finetuning/       # 16 papers - DeepSeek R1, GRPO, DPO
├── robotics/            #  6 papers - Eureka, Voyager, GR00T
├── security/            # 10 papers - Prompt injection, red-teaming
├── theorem-proving/     #  9 papers - LeanDojo, AlphaGeometry
└── web-agents/          #  7 papers - WebArena, Mind2Web

slides/                  # 92 presentation decks (504 MB)

---

Table of Contents

• Inference-Time Techniques
• Post-Training & Alignment
• Memory & Planning
• Agent Frameworks
• Code Generation & Software Agents
• Web & Multimodal Agents
• Enterprise & Workflow Agents
• Mathematics & Theorem Proving
• Robotics & Embodied Agents
• Scientific Discovery
• Safety & Security
• Evaluation & Benchmarking
• Neural & Symbolic Reasoning
• Agentic Reasoning & RL Fine-Tuning
• Agentic Architectures & Coordination
• Deep Reinforcement Learning

---

Inference-Time Techniques

Paper	Slides	Code	Media
Large Language Models as Optimizers	CS839 Prompting II	GitHub	🖼️
Large Language Models Cannot Self-Correct Reasoning Yet	-	-	🎨 🖼️
Teaching Large Language Models to Self-Debug	-	-	🎨 🖼️ 🎧
Chain-of-Thought Reasoning Without Prompting	CoT Princeton Lecture, CoT Toronto, CoT SJTU, CoT Interpretable ML, Concise CoT	GitHub (unofficial)	🎨 🖼️
Premise Order Matters in Reasoning with LLMs	-	-	🎨 🖼️
Chain-of-Thought Empowers Transformers	CoT Slides	-	🎨 🖼️

Post-Training & Alignment

Paper	Slides	Code	Media
Direct Preference Optimization (DPO)	DPO CMU, DPO UT Austin, DPO Toronto, DPO Jinen	GitHub	🎨 🖼️
Iterative Reasoning Preference Optimization	-	-	🎨 🖼️
Chain-of-Verification Reduces Hallucination	-	GitHub (unofficial)	🎨 🖼️
Unpacking DPO and PPO	DPO Slides	GitHub	🖼️
RLHF Background	RLHF UT Austin	-	-

Memory & Planning

Paper	Slides	Code	Media
Grokked Transformers are Implicit Reasoners	-	GitHub	🎨 🖼️
HippoRAG: Neurobiologically Inspired Long-Term Memory	HippoRAG NeurIPS	GitHub	🎨 🖼️
Is Your LLM Secretly a World Model of the Internet	-	GitHub	🖼️
Tree Search for Language Model Agents	-	GitHub	🖼️

Agent Frameworks

Paper	Slides	Code	Media
ReAct: Synergizing Reasoning and Acting	ReAct UVA Lecture	GitHub	🎨 🖼️ 🎧
AutoGen: Multi-Agent Conversation	-	GitHub	🎨 🖼️ 🎧
StateFlow: Enhancing LLM Task-Solving	-	GitHub	🖼️ 🎧
DSPy: Compiling Declarative Language Model	-	GitHub	🎨 🖼️ 🎧
LLM Agents Tutorials	EMNLP 2024 Tutorial, WWW 2024 Tutorial, Berkeley Training Agents	-	-

Code Generation & Software Agents

Paper	Slides	Code	Media
SWE-agent: Agent-Computer Interfaces	Software Agents (Neubig)	GitHub	🎨 🖼️
OpenHands: AI Software Developers	Software Agents (Neubig)	GitHub	🖼️ 🎧
Interactive Tools Assist LM Agents Security Vulnerabilities	Code Agents & Vulnerability Detection	GitHub	-
Big Sleep: LLM Vulnerabilities Real-World	Code Agents & Vulnerability Detection	-	-
SWE-bench Verified	-	GitHub	🖼️ 🎧

Web & Multimodal Agents

Paper	Slides	Code	Media
WebShop: Scalable Real-World Web Interaction	Multimodal Agents Berkeley	GitHub	-
Mind2Web: Generalist Agent for the Web	Multimodal Agents Berkeley	GitHub	-
WebArena: Realistic Web Environment	Multimodal Agents Berkeley, Web Agent Evaluation	GitHub	-
VisualWebArena	Multimodal Agents Berkeley	GitHub	-
AGUVIS: Unified Pure Vision Agents GUI	-	GitHub	-
BrowseComp: Web Browsing Benchmark	-	GitHub	-

Enterprise & Workflow Agents

Paper	Slides	Code	Media
WorkArena: Common Knowledge Work Tasks	-	GitHub	🖼️ 🎧
WorkArena++: Compositional Planning	-	GitHub	🖼️ 🎧
TapeAgents: Holistic Framework Agent Development	TapeAgents Slides	GitHub	🖼️ 🎧

Mathematics & Theorem Proving

Paper	Slides	Code	Media
LeanDojo: Theorem Proving Retrieval-Augmented	LeanDojo AITP, LeanDojo NeurIPS, Theorem Proving ML	GitHub	🎨
Autoformalization with Large Language Models	-	-	🎨
Autoformalizing Euclidean Geometry	-	GitHub	🎨
Draft, Sketch and Prove: Formal Theorem Provers	Theorem Proving ML	GitHub	🎨
miniCTX: Neural Theorem Proving Long-Contexts	-	GitHub	🎨
Lean-STaR: Interleave Thinking and Proving	Berkeley Slides	GitHub Website	🎨
ImProver: Agent-Based Automated Proof Optimization	-	GitHub	🎨
In-Context Learning Agent Formal Theorem-Proving	-	GitHub	-
Symbolic Regression: Learned Concept Library	-	GitHub	🖼️
AlphaGeometry: Solving Olympiad Geometry	-	GitHub	-

Robotics & Embodied Agents

Paper	Slides	Code	Media
Voyager: Open-Ended Embodied Agent	Voyager UT Austin	GitHub	🎨
Eureka: Human-Level Reward Design	Eureka Paper/Slides	GitHub	🎨 🖼️
DrEureka: Language Model Guided Sim-To-Real	-	GitHub	🎨 🖼️
Gran Turismo: Deep Reinforcement Learning	-	-	🖼️
GR00T N1: Foundation Model Humanoid	-	GitHub	🎨 🖼️
SLAC: Simulation-Pretrained Latent Action	-	-	-

Scientific Discovery

Paper	Slides	Code	Media
Paper2Agent: Research Papers as AI Agents	-	GitHub	🖼️ 🎧
OpenScholar: Synthesizing Scientific Literature	-	GitHub	🖼️

Safety & Security

Paper	Slides	Code	Media
DataSentinel: Game-Theoretic Detection Prompt Injection	Prompt Injection Duke	GitHub	-
AgentPoison: Red-teaming LLM Agents	Prompt Injection Duke	GitHub	🎨
Progent: Programmable Privilege Control	-	-	-
DecodingTrust: Trustworthiness GPT Models	-	GitHub	-
Representation Engineering: AI Transparency	-	GitHub	-
Extracting Training Data from LLMs	-	-	-
The Secret Sharer: Unintended Memorization	-	-	-
Privtrans: Privilege Separation	-	-	-

Evaluation & Benchmarking

Paper	Slides	Code	Media
Survey: Evaluation LLM-based Agents	AgentBench Multi-Turn NeurIPS	-	🖼️ 🎧
Adding Error Bars to Evals	-	GitHub	🖼️ 🎧
Tau2-Bench: Conversational Agents Dual-Control	-	GitHub	🖼️ 🎧
Data Science Agents	Data Science Agents Benchmark	-	-

Neural & Symbolic Reasoning

Paper	Slides	Code	Media
Beyond A-Star: Better Planning Transformers	-	GitHub	🖼️
Dualformer: Controllable Fast and Slow Thinking	-	GitHub	🖼️
Composing Global Optimizers: Algebraic Objects	-	-	🖼️
SurCo: Learning Linear Surrogates	-	-	🖼️

Agentic Reasoning & RL Fine-Tuning

Source: redhat-et/agentic-reasoning-reinforcement-fine-tuning

DeepSeek R1 & Reasoning Models

Paper	Slides	Code	Media
DeepSeek-R1: Reasoning via RL	DeepSeek R1 Intro, DeepSeek R1 Toronto, DeepSeek R1 CMU, DeepSeek R1 Seoul	GitHub	🎨 🖼️
DeepSeek R1: Implications for AI	DeepSeek R1 Intro	-	🎨 🖼️
DeepSeek R1: Are Reasoning Models Faithful?	-	-	🎨 🖼️
OpenAI O1 Replication Journey	-	GitHub	🎨 🖼️
Qwen QwQ Reasoning Model	-	HuggingFace	🎨 🖼️
Sky-T1: Training Small Reasoning LLMs	-	GitHub	🖼️
s1: Simple Test-Time Scaling	-	GitHub	🖼️

GRPO & RL Fine-Tuning

Paper	Slides	Code	Media
DeepSeekMath: GRPO Algorithm	Stanford RL for Reasoning	GitHub	🎨 🖼️
Guided GRPO: Adaptive Guidance	PTA-GRPO Planning	GitHub	🖼️
R-Search: Multi-Step Reasoning	Stanford RL for Reasoning	GitHub	🖼️
RL Fine-tuning: Instruction Following	-	-	🖼️
RFT Powers Multimodal Reasoning	-	-	🖼️
STILL-2: Distilling Reasoning	-	-	🖼️

Agentic RL

Paper	Slides	Code	Media
WebAgent-R1: Multi-Turn RL for Web Agents	-	GitHub	-
ARTIST: Agentic Reasoning & Tool Integration	ARTIST Microsoft	GitHub	🎨 🖼️ 🎧

Agentic Architectures & Coordination

Papers on multi-agent systems, decentralized coordination, and agentic frameworks

Decentralized Multi-Agent Systems

Paper	Slides	Code	Media
AgentNet: Decentralized Multi-Agent Coordination	-	GitHub	🎨 🖼️
MasRouter: Multi-Agent Routing	MasRouter ACL 2025	GitHub	🎨 🖼️
Multi-Agent RL Overview	Edinburgh MARL Intro	-	-

Device & Computer Control

Paper	Slides	Code	Media
DigiRL: Device Control Agents	DigiRL NeurIPS 2024	GitHub	🖼️ 🎧
OSWorld: Multimodal Agents Benchmark	-	GitHub	🖼️
OS-Harm: Computer Use Safety	OS-Harm Benchmark	GitHub	🖼️ 🎧

Agent Fine-Tuning & Tool Use

Paper	Slides	Code	Media
FireAct: Language Agent Fine-tuning	LLM Agents Tool Learning	GitHub	🖼️ 🎧
DeepSeek Janus Pro: Multimodal	-	GitHub	🎨 🖼️
PTA-GRPO: High-Level Planning	PTA-GRPO Planning	-	-
Stanford RL for Agents	Stanford RL Agents 2025	-	-
CMU LM Agents	CMU Language Models as Agents	-	-
Mannheim Tool Use	Mannheim LLM Agents Tool Use	-	-

Enterprise & Industry Guides

Resource	Description	Code
Intel AI Agents Architecture	AI agents resource guide	-
Cisco Agentic Frameworks	Overview of agentic frameworks	-

---

Deep Reinforcement Learning

See Full Deep RL Resources Guide - Comprehensive collection with 100+ resources and 92 slides

Value-Based Methods (DQN Family)

Paper	arXiv	Slides	Code	Media
Playing Atari with Deep RL (DQN)	1312.5602	CMU, CVUT, NTHU, Waterloo	OpenAI Baselines	-
Deep RL with Double Q-learning	1509.06461	CMU DQN	OpenAI Baselines	-
Dueling Network Architectures	1511.06581	Buffalo	OpenAI Baselines	-
Prioritized Experience Replay	1511.05952	Buffalo, Julien Vitay, ICML 2020	OpenAI Baselines	-
Rainbow: Combining Improvements	1710.02298	Prague, Berkeley, Wisconsin	Dopamine	-

Policy Gradient Methods

Paper	arXiv	Slides	Code	Media
Policy Gradient Methods	-	Toronto, Berkeley CS285, REINFORCE Stanford	Stable-Baselines3	-
Proximal Policy Optimization (PPO)	1707.06347	Waterloo, NTU Taiwan	OpenAI Baselines	-
Trust Region Policy Optimization (TRPO)	1502.05477	FAU, UT Austin, CMU Natural PG, Toronto PAIR	OpenAI Baselines	-
High-Dimensional Continuous Control (GAE)	1506.02438	Berkeley CS285	OpenAI Baselines	-

Actor-Critic Methods

Paper	arXiv	Slides	Code	Media
Asynchronous Methods (A3C)	1602.01783	WPI, Buffalo, NTU, UIUC, Julien Vitay	OpenAI Baselines	-
Continuous Control (DDPG)	1509.02971	Paderborn, FAU, Julien Vitay, Buffalo	Stable-Baselines3	-
Addressing Function Approximation (TD3)	1802.09477	Prague	Stable-Baselines3	-
Soft Actor-Critic (SAC)	1801.01290	Toronto PAIR, Purdue, Stanford CS231n, Prague	Stable-Baselines3	-

Temporal Difference & Q-Learning

Paper	arXiv	Slides	Code	Media
TD Learning Fundamentals	-	CMU, Michigan, Sutton & Barto	-	-
Q-Learning	-	Northeastern, CMU TD	-	-

Model-Based RL

Paper	arXiv	Slides	Code	Media
Model-Based RL	-	FAU, Toronto, Berkeley, CMU	MBRL-Lib	-

Imitation & Inverse RL

Paper	arXiv	Slides	Code	Media
Imitation Learning	-	WPI, EPFL	imitation	-
Inverse Reinforcement Learning	-	TU Darmstadt, Berkeley CS285	imitation	-

Introductory Lectures

Topic	Slides
Deep RL Introduction	Berkeley CS294, Berkeley 2017

Frameworks & Tools

Tool	Link	Description
OpenAI Gym	GitHub	RL environments
Gymnasium	GitHub	Maintained fork of Gym
Stable-Baselines3	GitHub	RL algorithms in PyTorch
Unity ML-Agents	GitHub	3D environments
PyTorch	pytorch.org	Deep learning framework
Google Dopamine	GitHub	RL research framework
CleanRL	GitHub	Single-file RL implementations
RLlib	GitHub	Scalable RL library

View all 100+ resources in DEEP_RL_RESOURCES.md

---

Recommended Study Path

Beginner

1. Start with WWW 2024 LLM Agents Tutorial - comprehensive overview
2. Read ReAct paper + slides + code
3. Study Chain-of-Thought with CoT Princeton Lecture

Intermediate

1. Software Agents (Neubig) for code agents + SWE-agent code
2. DPO CMU Lecture for alignment + DPO code
3. Multimodal Agents Berkeley for web agents + WebArena code

Advanced

1. LeanDojo slides for theorem proving + code
2. HippoRAG NeurIPS for memory systems + code
3. Prompt Injection Duke for security

Reasoning & RL Fine-Tuning Path

1. DeepSeek-R1 paper + DeepSeek R1 CMU slides + code
2. DeepSeekMath GRPO + Stanford RL for Reasoning + code
3. ARTIST paper for agentic reasoning with tools

---

License

Papers are property of their respective authors. This collection is for educational purposes.