Conference schedule, top papers, and analysis of the data for NeurIPS 2023!
Caption: Wordcloud of all NeurIPS 2023 titles
Welcome to the hub for all things NeurIPS 2023! We scraped the data for all 3500+ NeurIPS projects and dove into the depths of Hugging Face, GitHub, LinkedIn, and Arxiv to pick out the most interesting content.
In this repo, you will find:
The raw data is included in this repo. If you have ideas for other interesting analyses, feel free to create an issue or submit a PR!
For now, insights are organized into the following categories:
🔍 For the data analysis itself, check out the Jupyter Notebook!
🔍 And check out the blog post synthesizing the results here.
The top 10 authors with the most papers at NeurIPS 2023 are:
There were 13,012 unique authors at NeurIPS 2023, up from 9913 at NeurIPS 2022.
This continues the exponential explosion of unique authors over the past decade.
22% of titles introduced an acronym, up from 18% at NeurIPS 2022.
Using a CLIP model, we zero-shot
classified/predicted the modality of focus for each paper based on its abstract.
The categories were ["vision", "text", "audio", "tabular", "time series", "multimodal"]
.
By far the biggest category was multimodal, with a count of 1296. However, the
CLIP model's inclination towards "multimodal" may be somewhat biased by trying
to partially fit other modalities — the words multi-modal
and multimodal
only
show up in 156 abstracts, and phrases like vision-language
and text-to-image
only appear a handful of times across the dataset.
Themes occurring frequently include:
Title | Paper | Code | Project Page | Hugging Face | Blog |
---|---|---|---|---|---|
An Inverse Scaling Law for CLIP Training | |||||
Augmenting Language Models with Long-Term Memory | |||||
Chameleon: Plug-and-Play Compositional Reasoning with Large Language Models | Project | Blog | |||
Cheap and Quick: Efficient Vision-Language Instruction Tuning for Large Language Models | Project | Blog | |||
DataComp: In search of the next generation of multimodal datasets | Project | Blog | |||
Direct Preference Optimization: Your Language Model is Secretly a Reward Model | Blog | ||||
DreamSim: Learning New Dimensions of Human Visual Similarity using Synthetic Data | Project | Blog | |||
Fine-Tuning Language Models with Just Forward Passes | Blog | ||||
Generating Images with Multimodal Language Models | Project | ||||
Holistic Evaluation of Text-To-Image Models | Project | Blog | |||
HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face | |||||
ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generation | Blog | ||||
InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning | Blog | ||||
Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena | |||||
LAMM: Multi-Modal Large Language Models and Applications as AI Agents | Project | ||||
LIMA: Less Is More for Alignment | Blog | ||||
LLM-Pruner: On the Structural Pruning of Large Language Models | |||||
LightZero: A Unified Benchmark for Monte Carlo Tree Search in General Sequential Decision Scenario | |||||
MVDiffusion: Enabling Holistic Multi-view Image Generation with Correspondence-Aware Diffusion | Project | ||||
MagicBrush: A Manually Annotated Dataset for Instruction-Guided Image Editing | Project | Blog | |||
Mathematical Capabilities of ChatGPT | Project | ||||
Michelangelo: Conditional 3D Shape Generation based on Shape-Image-Text Aligned Latent Representation | Project | ||||
Motion-X: A Large-scale 3D Expressive Whole-body Human Motion Dataset | Project | ||||
MotionGPT: Human Motion as Foreign Language | Project | Blog | |||
OBELICS: An Open Web-Scale Filtered Dataset of Interleaved Image-Text Documents | Blog | ||||
Photoswap: Personalized Subject Swapping in Images | Project | ||||
Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation | Blog | ||||
QLoRA: Efficient Finetuning of Quantized LLMs | Blog | ||||
Reflexion: Language Agents with Verbal Reinforcement Learning | Blog | ||||
ResShift: Efficient Diffusion Model for Image Super-resolution by Residual Shifting | Project | Blog | |||
Segment Anything in 3D with NeRFs | Project | Blog | |||
Segment Anything in High Quality | Blog | ||||
Segment Everything Everywhere All at Once | |||||
Self-Refine: Iterative Refinement with Self-Feedback | Project | Blog | |||
Simple and Controllable Music Generation | Blog | ||||
Squeeze, Recover and Relabel: Dataset Condensation at ImageNet Scale From A New Perspective | |||||
The RefinedWeb Dataset for Falcon LLM: Outperforming Curated Corpora with Web Data, and Web Data Only | Blog | ||||
Toolformer: Language Models Can Teach Themselves to Use Tools | Blog | ||||
Unlimiformer: Long-Range Transformers with Unlimited Length Input | Blog | ||||
Visual Instruction Tuning | Project | Blog |
Note: GitHub automatically truncates files larger than 512 KB. To have all papers display on GitHub, we've split the file up by session.