Paper Graph

Browse by domain

Each domain is a sub-collection of papers, sorted newest first. Tap a card to open the timeline for that domain.

• 

  cs lg

  2723 papers —

• 

  cs cv

  2231 papers —

• 

  cs cl

  1694 papers —

• 

  Training Methods

  821 papers — Optimizers, schedulers, and training infrastructure.

• 

  Multimodal

  529 papers — Vision-language, audio, and multi-modal fusion.

• 

  stat ml

  471 papers —

• 

  cs ai

  414 papers —

• 

  cs ir

  400 papers —

• 

  Vision

  392 papers — Computer vision and ViT.

• 

  Inference Optimization

  391 papers — Speculative decoding, quantization at inference, and runtime tricks.

• 

  RL Training

  368 papers — RLHF, GRPO, RLVR, and reward modeling.

• 

  Architecture

  337 papers — Transformers, attention variants, and model design.

• 

  Reasoning

  334 papers — CoT, chain-of-thought, test-time compute.

• 

  Pretraining

  322 papers — Base model pretraining recipes, scaling laws, and corpora.

• 

  Evaluation

  321 papers — Benchmarks, leaderboards, and eval methodologies.

• 

  Agents

  306 papers — Agentic systems, tool use, and autonomous reasoning.

• 

  cs ds

  263 papers —

• 

  eess as

  248 papers —

• 

  cs ro

  228 papers —

• 

  cs hc

  224 papers —

• 

  LLM Systems

  213 papers — End-to-end systems for serving and orchestrating LLMs.

• 

  cs gt

  211 papers —

• 

  Data

  205 papers — Data curation and dataset construction.

• 

  Serving

  172 papers — vLLM, SGLang, TensorRT-LLM, and inference servers.

• 

  Diffusion

  170 papers — Diffusion models and generative processes.

• 

  Alignment

  166 papers — RLHF, DPO, preference learning, and safety tuning.

• 

  cs si

  148 papers —

• 

  cs cr

  146 papers —

• 

  eess iv

  146 papers —

• 

  cs sd

  134 papers —

• 

  cs se

  126 papers —

• 

  cs dc

  125 papers —

• 

  Context Optimization

  124 papers — KV cache compression, long context, and efficient attention.

• 

  cs db

  115 papers —

• 

  cs cy

  112 papers —

• 

  Safety

  107 papers — Jailbreak resistance, alignment, and red-teaming.

• 

  cs it

  107 papers —

• 

  Low Precision

  101 papers — FP8, FP4, quantization, and mixed-precision training.

• 

  cs gr

  97 papers —

• 

  Retrieval

  93 papers — Embeddings, RAG, and vector retrieval.

• 

  Scaling Laws

  89 papers — Compute/data/parameter scaling, Chinchilla, and compute-optimal training.

• 

  Mixture of Experts

  89 papers — MoE routing, sparsity, and expert architectures.

• 

  cs pl

  86 papers —

• 

  Distributed Training

  84 papers — Multi-GPU and multi-node training systems.

• 

  Code

  78 papers — Code generation and code-LLM systems.

• 

  stat me

  74 papers —

• 

  Uncategorized

  71 papers — Papers awaiting categorization.

• 

  cs ne

  64 papers —

• 

  Prompting

  59 papers — Prompt engineering, CoT, in-context learning.

• 

  cs lo

  45 papers —

• 

  cs cc

  45 papers —

• 

  cs ni

  43 papers —

• 

  cs ma

  41 papers —

• 

  stat ap

  37 papers —

• 

  cs mm

  36 papers —

• 

  cs ar

  33 papers —

• 

  eess sp

  30 papers —

• 

  eess sy

  28 papers —

• 

  cs cg

  19 papers —

• 

  stat co

  14 papers —

• 

  cs dl

  14 papers —

• 

  cs dm

  9 papers —

• 

  cs ce

  9 papers —

• 

  cs ms

  8 papers —

• 

  cs pf

  7 papers —

• 

  cs et

  6 papers —

• 

  cs fl

  4 papers —

• 

  cs oh

  3 papers —

• 

  cs gl

  1 paper —

• 

  stat ot

  1 paper —

• 

  cs os

  1 paper —

• 

  math st

  1 paper —

• 

  cs sc

  1 paper —

Your progress

Years covered

Topic threads

Cross-domain threads of papers that share a topic — RLHF, MoE, KV caches, agents, and more.