NeurIPS 2023 accepted papers on Large language models

Language, safety and interpretability · 82 papers

• Toolformer: Language Models Can Teach Themselves to Use Tools (507 citations)
• VisionLLM: Large Language Model is also an Open-Ended Decoder for Vision-Centric Tasks (132 citations)
• DIN-SQL: Decomposed In-Context Learning of Text-to-SQL with Self-Correction (131 citations)
• Jailbroken: How Does LLM Safety Training Fail? (130 citations)
• What can Large Language Models do in chemistry? A comprehensive benchmark on eight tasks (92 citations)
• Paraphrasing evades detectors of AI-generated text, but retrieval is an effective defense (90 citations)
• Can LLM Already Serve as A Database Interface? A BIg Bench for Large-Scale Database Grounded Text-to-SQLs (87 citations)
• Large Language Model as Attributed Training Data Generator: A Tale of Diversity and Bias (72 citations)
• 3D-LLM: Injecting the 3D World into Large Language Models (66 citations)
• In-Context Impersonation Reveals Large Language Models' Strengths and Biases (36 citations)
• Intrinsic Dimension Estimation for Robust Detection of AI-Generated Texts (31 citations)
• RADAR: Robust AI-Text Detection via Adversarial Learning (29 citations)

and 70 more in the interactive map.

Agents, code and math · 39 papers

• Self-Refine: Iterative Refinement with Self-Feedback (350 citations)
• Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generation (172 citations)
• LegalBench: A Collaboratively Built Benchmark for Measuring Legal Reasoning in Large Language Models (151 citations)
• C-Eval: A Multi-Level Multi-Discipline Chinese Evaluation Suite for Foundation Models (92 citations)
• Faith and Fate: Limits of Transformers on Compositionality (70 citations)
• On the Planning Abilities of Large Language Models - A Critical Investigation (53 citations)
• PlanBench: An Extensible Benchmark for Evaluating Large Language Models on Planning and Reasoning about Change (46 citations)
• Large Language Models as Commonsense Knowledge for Large-Scale Task Planning (40 citations)
• LeanDojo: Theorem Proving with Retrieval-Augmented Language Models (39 citations)
• Large Language Models for Automated Data Science: Introducing CAAFE for Context-Aware Automated Feature Engineering (38 citations)
• Leveraging Pre-trained Large Language Models to Construct and Utilize World Models for Model-based Task Planning (33 citations)
• M3Exam: A Multilingual, Multimodal, Multilevel Benchmark for Examining Large Language Models (32 citations)

and 27 more in the interactive map.

Retrieval and knowledge · 24 papers

• ToolQA: A Dataset for LLM Question Answering with External Tools (39 citations)
• Augmenting Language Models with Long-Term Memory (33 citations)
• RealTime QA: What's the Answer Right Now? (28 citations)
• Can Language Models Solve Graph Problems in Natural Language? (27 citations)
• Lift Yourself Up: Retrieval-augmented Text Generation with Self-Memory (20 citations)
• FELM: Benchmarking Factuality Evaluation of Large Language Models (12 citations)
• AVeriTeC: A Dataset for Real-world Claim Verification with Evidence from the Web (7 citations)
• Evaluating Open-QA Evaluation (5 citations)
• RECKONING: Reasoning through Dynamic Knowledge Encoding (5 citations)
• CARE-MI: Chinese Benchmark for Misinformation Evaluation in Maternity and Infant Care (4 citations)
• Propagating Knowledge Updates to LMs Through Distillation (4 citations)
• Scientific Document Retrieval using Multi-level Aspect-based Queries (4 citations)

and 12 more in the interactive map.

AI and society · 23 papers

• Quantus: An Explainable AI Toolkit for Responsible Evaluation of Neural Network Explanations and Beyond (115 citations)
• OpenAGI: When LLM Meets Domain Experts (74 citations)
• DataPerf: Benchmarks for Data-Centric AI Development (50 citations)
• ClimateLearn: Benchmarking Machine Learning for Weather and Climate Modeling (9 citations)
• Reimagining Synthetic Tabular Data Generation through Data-Centric AI: A Comprehensive Benchmark (8 citations)
• The Cambridge Law Corpus: A Dataset for Legal AI Research (8 citations)
• Scale Alone Does not Improve Mechanistic Interpretability in Vision Models (6 citations)
• trajdata: A Unified Interface to Multiple Human Trajectory Datasets (6 citations)
• ClimateSet: A Large-Scale Climate Model Dataset for Machine Learning (5 citations)
• Data Portraits: Recording Foundation Model Training Data (5 citations)
• DICES Dataset: Diversity in Conversational AI Evaluation for Safety (5 citations)
• LithoBench: Benchmarking AI Computational Lithography for Semiconductor Manufacturing (5 citations)

and 11 more in the interactive map.

Reasoning and chain of thought · 19 papers

• Tree of Thoughts: Deliberate Problem Solving with Large Language Models (566 citations)
• Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting (107 citations)
• Language Models can Solve Computer Tasks (67 citations)
• H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models (66 citations)
• Inference-Time Intervention: Eliciting Truthful Answers from a Language Model (56 citations)
• Guiding Large Language Models via Directional Stimulus Prompting (28 citations)
• Towards Revealing the Mystery behind Chain of Thought: A Theoretical Perspective (27 citations)
• DDCoT: Duty-Distinct Chain-of-Thought Prompting for Multimodal Reasoning in Language Models (21 citations)
• Response Length Perception and Sequence Scheduling: An LLM-Empowered LLM Inference Pipeline (15 citations)
• Why think step by step? Reasoning emerges from the locality of experience (13 citations)
• Self-Evaluation Guided Beam Search for Reasoning (10 citations)
• Evaluating and Improving Tool-Augmented Computation-Intensive Math Reasoning (9 citations)

and 7 more in the interactive map.

Instruction and fine-tuning · 17 papers

• How Far Can Camels Go? Exploring the State of Instruction Tuning on Open Resources (38 citations)
• GPT4Tools: Teaching Large Language Model to Use Tools via Self-instruction (32 citations)
• Memory-Efficient Fine-Tuning of Compressed Large Language Models via sub-4-bit Integer Quantization (27 citations)
• The Goldilocks of Pragmatic Understanding: Fine-Tuning Strategy Matters for Implicature Resolution by LLMs (21 citations)
• To Repeat or Not To Repeat: Insights from Scaling LLM under Token-Crisis (20 citations)
• Distributed Inference and Fine-tuning of Large Language Models Over The Internet (17 citations)
• On the Exploitability of Instruction Tuning (13 citations)
• Language Models are Weak Learners (12 citations)
• LayoutPrompter: Awaken the Design Ability of Large Language Models (10 citations)
• PIXIU: A Comprehensive Benchmark, Instruction Dataset and Large Language Model for Finance (10 citations)
• Revisiting Out-of-distribution Robustness in NLP: Benchmarks, Analysis, and LLMs Evaluations (7 citations)
• Don’t Stop Pretraining? Make Prompt-based Fine-tuning Powerful Learner (6 citations)

and 5 more in the interactive map.

Alignment and preferences · 11 papers

• Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena (810 citations)
• Direct Preference Optimization: Your Language Model is Secretly a Reward Model (418 citations)
• OpenAssistant Conversations - Democratizing Large Language Model Alignment (130 citations)
• LIMA: Less Is More for Alignment (128 citations)
• Principle-Driven Self-Alignment of Language Models from Scratch with Minimal Human Supervision (63 citations)
• AlpacaFarm: A Simulation Framework for Methods that Learn from Human Feedback (54 citations)
• BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset (35 citations)
• Fine-Grained Human Feedback Gives Better Rewards for Language Model Training (34 citations)
• RRHF: Rank Responses to Align Language Models with Human Feedback (15 citations)
• Aligning Language Models with Human Preferences via a Bayesian Approach (1 citation)
• Preference-grounded Token-level Guidance for Language Model Fine-tuning (1 citation)