└── README.md /README.md: -------------------------------------------------------------------------------- 1 | # Elicit Machine Learning Reading List 2 | 3 | ## Purpose 4 | 5 | The purpose of this curriculum is to help new [Elicit](https://elicit.com/) employees learn background in machine learning, with a focus on language models. I’ve tried to strike a balance between papers that are relevant for deploying ML in production and techniques that matter for longer-term scalability. 6 | 7 | If you don’t work at Elicit yet - we’re [hiring ML and software engineers](https://elicit.com/careers). 8 | 9 | ## How to read 10 | 11 | Recommended reading order: 12 | 13 | 1. Read “Tier 1” for all topics 14 | 2. Read “Tier 2” for all topics 15 | 3. Etc 16 | 17 | ✨ = Added after 2025/11/26 18 | 19 | ## Table of contents 20 | 21 | - [Fundamentals](#fundamentals) 22 | - [Introduction to machine learning](#introduction-to-machine-learning) 23 | - [Transformers](#transformers) 24 | - [Key foundation model architectures](#key-foundation-model-architectures) 25 | - [Training and finetuning](#training-and-finetuning) 26 | - [Reasoning and runtime strategies](#reasoning-and-runtime-strategies) 27 | - [In-context reasoning](#in-context-reasoning) 28 | - [Task decomposition](#task-decomposition) 29 | - [Debate](#debate) 30 | - [Tool use and scaffolding](#tool-use-and-scaffolding) 31 | - [Honesty, factuality, and epistemics](#honesty-factuality-and-epistemics) 32 | - [Applications](#applications) 33 | - [Science](#science) 34 | - [Forecasting](#forecasting) 35 | - [Search and ranking](#search-and-ranking) 36 | - [ML in practice](#ml-in-practice) 37 | - [Production deployment](#production-deployment) 38 | - [Benchmarks](#benchmarks) 39 | - [Datasets](#datasets) 40 | - [Advanced topics](#advanced-topics) 41 | - [World models and causality](#world-models-and-causality) 42 | - [Planning](#planning) 43 | - [Uncertainty, calibration, and active learning](#uncertainty-calibration-and-active-learning) 44 | - [Interpretability and model editing](#interpretability-and-model-editing) 45 | - [Reinforcement learning](#reinforcement-learning) 46 | - [The big picture](#the-big-picture) 47 | - [AI scaling](#ai-scaling) 48 | - [AI safety](#ai-safety) 49 | - [Economic and social impacts](#economic-and-social-impacts) 50 | - [Philosophy](#philosophy) 51 | - [Maintainer](#maintainer) 52 | 53 | ## Fundamentals 54 | 55 | ### Introduction to machine learning 56 | 57 | **Tier 1** 58 | 59 | - [A short introduction to machine learning](https://www.alignmentforum.org/posts/qE73pqxAZmeACsAdF/a-short-introduction-to-machine-learning) 60 | - [But what is a neural network?](https://www.youtube.com/watch?v=aircAruvnKk&t=0s) 61 | - [Gradient descent, how neural networks learn](https://www.youtube.com/watch?v=IHZwWFHWa-w) 62 | 63 | **Tier 2** 64 | 65 | - [An intuitive understanding of backpropagation](https://cs231n.github.io/optimization-2/) 66 | - [What is backpropagation really doing?](https://www.youtube.com/watch?v=Ilg3gGewQ5U&list=PLZHQObOWTQDNU6R1_67000Dx_ZCJB-3pi&index=4) 67 | - [An introduction to deep reinforcement learning](https://thomassimonini.medium.com/an-introduction-to-deep-reinforcement-learning-17a565999c0c) 68 | 69 | **Tier 3** 70 | 71 | - [The spelled-out intro to neural networks and backpropagation: building micrograd](https://www.youtube.com/watch?v=VMj-3S1tku0) (Karpathy) 72 | - [Backpropagation calculus](https://www.youtube.com/watch?v=tIeHLnjs5U8&list=PLZHQObOWTQDNU6R1_67000Dx_ZCJB-3pi&index=5) 73 | 74 | ### Transformers 75 | 76 | **Tier 1** 77 | 78 | - ✨ [Intro to Large Language Models](https://www.youtube.com/watch?v=zjkBMFhNj_g) (Karpathy) 79 | - [But what is a GPT? Visual intro to transformers](https://www.youtube.com/watch?v=wjZofJX0v4M) 80 | - [Attention in transformers, visually explained](https://www.youtube.com/watch?v=eMlx5fFNoYc) 81 | - [Attention? Attention!](https://lilianweng.github.io/posts/2018-06-24-attention/) 82 | - [The Illustrated Transformer](http://jalammar.github.io/illustrated-transformer/) 83 | 84 | **Tier 2** 85 | 86 | - ✨ [Deep Dive into LLMs like ChatGPT](https://www.youtube.com/watch?v=7xTGNNLPyMI) (Karpathy) 87 | - [Let's build the GPT Tokenizer](https://www.youtube.com/watch?v=zduSFxRajkE) (Karpathy) 88 | - [The Illustrated GPT-2 (Visualizing Transformer Language Models)](https://jalammar.github.io/illustrated-gpt2/) 89 | - [Neural Machine Translation by Jointly Learning to Align and Translate](https://arxiv.org/pdf/1409.0473) 90 | - [Attention Is All You Need](https://arxiv.org/abs/1706.03762) 91 | 92 | **Tier 3** 93 | 94 | - ✨ [The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A"](https://arxiv.org/abs/2309.12288) 95 | - [The Annotated Transformer](https://nlp.seas.harvard.edu/2018/04/03/attention.html) 96 | - [TabPFN: A Transformer That Solves Small Tabular Classification Problems in a Second](https://arxiv.org/abs/2207.01848) 97 | - [Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets](https://arxiv.org/abs/2201.02177) 98 | - [A Mathematical Framework for Transformer Circuits](https://transformer-circuits.pub/2021/framework/index.html) 99 | 100 |
Tier 4+ 101 | 102 | - [A Practical Survey on Faster and Lighter Transformers](https://arxiv.org/abs/2103.14636) 103 | - [Compositional Capabilities of Autoregressive Transformers: A Study on Synthetic, Interpretable Tasks](https://arxiv.org/abs/2311.12997) 104 | - [Memorizing Transformers](https://arxiv.org/abs/2203.08913) 105 | - [Transformer Feed-Forward Layers Are Key-Value Memories](https://arxiv.org/abs/2012.14913) 106 | 107 |
108 | 109 | ### Key foundation model architectures 110 | 111 | **Tier 1** 112 | 113 | - [Language Models are Unsupervised Multitask Learners](https://www.semanticscholar.org/paper/Language-Models-are-Unsupervised-Multitask-Learners-Radford-Wu/9405cc0d6169988371b2755e573cc28650d14dfe) (GPT-2) 114 | - [Language Models are Few-Shot Learners](https://arxiv.org/abs/2005.14165) (GPT-3) 115 | 116 | **Tier 2** 117 | 118 | - ✨ [DeepSeek-R1](https://arxiv.org/abs/2501.12948) (DeepSeek-R1) 119 | - ✨ [DeepSeek-V3 Technical Report](https://arxiv.org/abs/2412.19437) (DeepSeek-V3) 120 | - ✨ [The Llama 3 Herd of Models](https://arxiv.org/abs/2407.21783) (Llama 3) 121 | - [LLaMA: Open and Efficient Foundation Language Models](http://arxiv.org/abs/2302.13971) (LLaMA) 122 | - [Training language models to follow instructions with human feedback](https://arxiv.org/abs/2203.02155) (OpenAI Instruct) 123 | 124 | **Tier 3** 125 | 126 | - ✨ [LLaMA 2: Open Foundation and Fine-Tuned Chat Models](https://arxiv.org/abs/2307.09288) (LLaMA 2) 127 | - ✨ [Qwen2.5 Technical Report](https://arxiv.org/abs/2412.15115) (Qwen2.5) 128 | - ✨ [Titans: Learning to Memorize at Test Time](https://arxiv.org/abs/2501.00663) 129 | - ✨ [Byte Latent Transformer](https://arxiv.org/abs/2412.09871) 130 | - ✨ [Phi-4 Technical Report](https://arxiv.org/abs/2412.08905) (phi-4) 131 | 132 |
Tier 4+ 133 | 134 | - [Evaluating Large Language Models Trained on Code](https://arxiv.org/abs/2107.03374) (OpenAI Codex) 135 | - [Mistral 7B](http://arxiv.org/abs/2310.06825) (Mistral) 136 | - [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://arxiv.org/abs/1910.10683) (T5) 137 | - [Gemini: A Family of Highly Capable Multimodal Models](https://storage.googleapis.com/deepmind-media/gemini/gemini_1_report.pdf) (Gemini) 138 | - [Mamba: Linear-Time Sequence Modeling with Selective State Spaces](https://arxiv.org/abs/2312.00752v1) (Mamba) 139 | - [Scaling Instruction-Finetuned Language Models](https://arxiv.org/abs/2210.11416) (Flan) 140 | - [Efficiently Modeling Long Sequences with Structured State Spaces](https://arxiv.org/abs/2111.00396) ([video](https://www.youtube.com/watch?v=EvQ3ncuriCM)) (S4) 141 | - [Consistency Models](http://arxiv.org/abs/2303.01469) 142 | - [Model Card and Evaluations for Claude Models](https://www-cdn.anthropic.com/bd2a28d2535bfb0494cc8e2a3bf135d2e7523226/Model-Card-Claude-2.pdf) (Claude 2) 143 | - [OLMo: Accelerating the Science of Language Models](http://arxiv.org/abs/2402.00838) 144 | - [PaLM 2 Technical Report](https://arxiv.org/abs/2305.10403) (Palm 2) 145 | - [Textbooks Are All You Need II: phi-1.5 technical report](http://arxiv.org/abs/2309.05463) (phi 1.5) 146 | - [Visual Instruction Tuning](http://arxiv.org/abs/2304.08485) (LLaVA) 147 | - [A General Language Assistant as a Laboratory for Alignment](https://arxiv.org/abs/2112.00861) 148 | - [Finetuned Language Models Are Zero-Shot Learners](https://arxiv.org/abs/2109.01652) (Google Instruct) 149 | - [Galactica: A Large Language Model for Science](https://arxiv.org/abs/2211.09085) 150 | - [LaMDA: Language Models for Dialog Applications](https://arxiv.org/abs/2201.08239) (Google Dialog) 151 | - [OPT: Open Pre-trained Transformer Language Models](https://arxiv.org/abs/2112.11446) (Meta GPT-3) 152 | - [PaLM: Scaling Language Modeling with Pathways](https://arxiv.org/abs/2204.02311) (PaLM) 153 | - [Program Synthesis with Large Language Models](https://arxiv.org/abs/2108.07732) (Google Codex) 154 | - [Scaling Language Models: Methods, Analysis & Insights from Training Gopher](https://arxiv.org/abs/2112.11446) (Gopher) 155 | - [Solving Quantitative Reasoning Problems with Language Models](https://arxiv.org/abs/2206.14858) (Minerva) 156 | - [UL2: Unifying Language Learning Paradigms](http://aima.cs.berkeley.edu/) (UL2) 157 | 158 |
159 | 160 | ### Training and finetuning 161 | 162 | **Tier 2** 163 | 164 | - ✨ [Deep Learning Tuning Playbook](https://github.com/google-research/tuning_playbook) 165 | - [Tensor Programs V: Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer](https://arxiv.org/abs/2203.03466) 166 | - [Learning to summarise with human feedback](https://arxiv.org/abs/2009.01325) 167 | - [Training Verifiers to Solve Math Word Problems](https://arxiv.org/abs/2110.14168) 168 | 169 | **Tier 3** 170 | 171 | - ✨ [Better & Faster Large Language Models via Multi-token Prediction](https://arxiv.org/abs/2404.19737) 172 | - ✨ [LoRA vs Full Fine-tuning: An Illusion of Equivalence](https://arxiv.org/abs/2410.21228) 173 | - ✨ [QLoRA: Efficient Finetuning of Quantized LLMs](http://arxiv.org/abs/2305.14314) 174 | - [Pretraining Language Models with Human Preferences](http://arxiv.org/abs/2302.08582) 175 | - [Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision](http://arxiv.org/abs/2312.09390) 176 | 177 |
Tier 4+ 178 | 179 | - [Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learning](https://arxiv.org/abs/2205.05638v1) 180 | - [Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Models](http://arxiv.org/abs/2312.06585) 181 | - [Improving Code Generation by Training with Natural Language Feedback](http://arxiv.org/abs/2303.16749) 182 | - [Language Modeling Is Compression](https://arxiv.org/abs/2309.10668v1) 183 | - [LIMA: Less Is More for Alignment](http://arxiv.org/abs/2305.11206) 184 | - [Learning to Compress Prompts with Gist Tokens](http://arxiv.org/abs/2304.08467) 185 | - [Lost in the Middle: How Language Models Use Long Contexts](http://arxiv.org/abs/2307.03172) 186 | - [LoRA: Low-Rank Adaptation of Large Language Models](https://arxiv.org/abs/2106.09685) 187 | - [Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking](http://arxiv.org/abs/2403.09629) 188 | - [Reinforced Self-Training (ReST) for Language Modeling](http://arxiv.org/abs/2308.08998) 189 | - [Solving olympiad geometry without human demonstrations](https://www.nature.com/articles/s41586-023-06747-5) 190 | - [Tell, don't show: Declarative facts influence how LLMs generalize](http://arxiv.org/abs/2312.07779) 191 | - [Textbooks Are All You Need](http://arxiv.org/abs/2306.11644) 192 | - [TinyStories: How Small Can Language Models Be and Still Speak Coherent English?](http://arxiv.org/abs/2305.07759) 193 | - [Training Language Models with Language Feedback at Scale](http://arxiv.org/abs/2303.16755) 194 | - [Turing Complete Transformers: Two Transformers Are More Powerful Than One](https://openreview.net/forum?id=MGWsPGogLH) 195 | - [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://arxiv.org/abs/2105.13626) 196 | - [Data Distributional Properties Drive Emergent In-Context Learning in Transformers](https://arxiv.org/abs/2205.05055) 197 | - [Diffusion-LM Improves Controllable Text Generation](https://arxiv.org/abs/2205.14217) 198 | - [ERNIE 3.0: Large-scale Knowledge Enhanced Pre-training for Language Understanding and Generation](https://arxiv.org/abs/2107.02137) 199 | - [Efficient Training of Language Models to Fill in the Middle](https://arxiv.org/abs/2207.14255) 200 | - [ExT5: Towards Extreme Multi-Task Scaling for Transfer Learning](https://arxiv.org/abs/2111.10952) 201 | - [Prefix-Tuning: Optimizing Continuous Prompts for Generation](https://arxiv.org/abs/2101.00190) 202 | - [Self-Attention Between Datapoints: Going Beyond Individual Input-Output Pairs in Deep Learning](https://arxiv.org/abs/2106.02584) 203 | - [True Few-Shot Learning with Prompts -- A Real-World Perspective](https://arxiv.org/abs/2111.13440) 204 | 205 |
206 | 207 | ## Reasoning and runtime strategies 208 | 209 | ### In-context reasoning 210 | 211 | **Tier 2** 212 | 213 | - ✨ [Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters](https://arxiv.org/abs/2408.03314) 214 | - [Chain of Thought Prompting Elicits Reasoning in Large Language Models](https://arxiv.org/abs/2201.11903) 215 | - [Large Language Models are Zero-Shot Reasoners](https://arxiv.org/abs/2205.11916) (Let's think step by step) 216 | - [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://arxiv.org/abs/2203.11171) 217 | 218 | **Tier 3** 219 | 220 | - ✨ [s1: Simple test-time scaling](https://arxiv.org/abs/2501.19393) 221 | - ✨ [Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRs](https://arxiv.org/abs/2503.01307) 222 | - ✨ [The Surprising Effectiveness of Test-Time Training for Abstract Reasoning](https://arxiv.org/abs/2411.07279) 223 | - ✨ [Large Language Models Cannot Self-Correct Reasoning Yet](https://arxiv.org/abs/2310.01798v1) 224 | - [Chain-of-Thought Reasoning Without Prompting](http://arxiv.org/abs/2402.10200) 225 | 226 |
Tier 4+ 227 | 228 | - [Why think step-by-step? Reasoning emerges from the locality of experience](http://arxiv.org/abs/2304.03843) 229 | - [Baldur: Whole-Proof Generation and Repair with Large Language Models](https://arxiv.org/abs/2303.04910v1) 230 | - [Bias-Augmented Consistency Training Reduces Biased Reasoning in Chain-of-Thought](http://arxiv.org/abs/2403.05518) 231 | - [Certified Reasoning with Language Models](http://arxiv.org/abs/2306.04031) 232 | - [Hypothesis Search: Inductive Reasoning with Language Models](http://arxiv.org/abs/2309.05660) 233 | - [LLMs and the Abstraction and Reasoning Corpus: Successes, Failures, and the Importance of Object-based Representations](http://arxiv.org/abs/2305.18354) 234 | - [Stream of Search (SoS): Learning to Search in Language](http://arxiv.org/abs/2404.03683) 235 | - [Training Chain-of-Thought via Latent-Variable Inference](http://arxiv.org/abs/2312.02179) 236 | - [Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?](https://arxiv.org/abs/2202.12837) 237 | - [Surface Form Competition: Why the Highest Probability Answer Isn’t Always Right](https://arxiv.org/abs/2104.08315) 238 | 239 |
240 | 241 | ### Task decomposition 242 | 243 | **Tier 1** 244 | 245 | - [Supervise Process, not Outcomes](https://ought.org/updates/2022-04-06-process) 246 | - [Supervising strong learners by amplifying weak experts](https://arxiv.org/abs/1810.08575) 247 | 248 | **Tier 2** 249 | 250 | - [Tree of Thoughts: Deliberate Problem Solving with Large Language Models](http://arxiv.org/abs/2305.10601) 251 | - [Factored cognition](https://ought.org/research/factored-cognition) 252 | - [Iterated Distillation and Amplification](https://ai-alignment.com/iterated-distillation-and-amplification-157debfd1616) 253 | - [Recursively Summarizing Books with Human Feedback](https://arxiv.org/abs/2109.10862) 254 | - [Solving math word problems with process-based and outcome-based feedback](https://arxiv.org/abs/2211.14275) 255 | 256 | **Tier 3** 257 | 258 | - [Factored Verification: Detecting and Reducing Hallucination in Summaries of Academic Papers](https://arxiv.org/abs/2310.10627) 259 | - [Faithful Reasoning Using Large Language Models](https://arxiv.org/abs/2208.14271) 260 | - [Iterated Decomposition: Improving Science Q&A by Supervising Reasoning Processes](https://arxiv.org/abs/2301.01751) 261 | - [Language Model Cascades](https://arxiv.org/abs/2207.10342) 262 | 263 |
Tier 4+ 264 | 265 | - [Decontextualization: Making Sentences Stand-Alone](https://doi.org/10.1162/tacl_a_00377) 266 | - [Factored Cognition Primer](https://primer.ought.org) 267 | - [Graph of Thoughts: Solving Elaborate Problems with Large Language Models](http://arxiv.org/abs/2308.09687) 268 | - [Parsel: A Unified Natural Language Framework for Algorithmic Reasoning](http://arxiv.org/abs/2212.10561) 269 | - [AI Chains: Transparent and Controllable Human-AI Interaction by Chaining Large Language Model Prompts](https://arxiv.org/abs/2110.01691) 270 | - [Challenging BIG-Bench tasks and whether chain-of-thought can solve them](https://arxiv.org/abs/2210.09261) 271 | - [Evaluating Arguments One Step at a Time](https://ought.org/updates/2020-01-11-arguments) 272 | - [Least-to-Most Prompting Enables Complex Reasoning in Large Language Models](https://arxiv.org/abs/2205.11822) 273 | - [Maieutic Prompting: Logically Consistent Reasoning with Recursive Explanations](https://arxiv.org/abs/2205.11822) 274 | - [Measuring and narrowing the compositionality gap in language models](https://arxiv.org/abs/2210.03350) 275 | - [PAL: Program-aided Language Models](https://arxiv.org/abs/2211.10435) 276 | - [ReAct: Synergizing Reasoning and Acting in Language Models](https://arxiv.org/abs/2210.03629) 277 | - [Selection-Inference: Exploiting Large Language Models for Interpretable Logical Reasoning](https://arxiv.org/abs/2205.10625) 278 | - [Show Your Work: Scratchpads for Intermediate Computation with Language Models](https://arxiv.org/abs/2112.00114) 279 | - [Summ^N: A Multi-Stage Summarization Framework for Long Input Dialogues and Documents](https://arxiv.org/abs/2110.10150) 280 | - [Thinksum: probabilistic reasoning over sets using large language models](https://arxiv.org/abs/2210.01293) 281 | 282 |
283 | 284 | ### Debate 285 | 286 | **Tier 2** 287 | 288 | - [AI safety via debate](https://openai.com/blog/debate/) 289 | 290 | **Tier 3** 291 | 292 | - ✨ [Avoiding Obfuscation with Prover-Estimator Debate](https://arxiv.org/abs/2506.13609) 293 | - ✨ [Improving Factuality and Reasoning in Language Models through Multiagent Debate](http://arxiv.org/abs/2305.14325) 294 | - ✨ [Prover-Verifier Games Improve Legibility of LLM Outputs](https://arxiv.org/abs/2407.13692) 295 | - [Debate Helps Supervise Unreliable Experts](https://twitter.com/joshua_clymer/status/1724851456967417872) 296 | 297 |
Tier 4+ 298 | 299 | - [Scalable AI Safety via Doubly-Efficient Debate](http://arxiv.org/abs/2311.14125) 300 | - [Two-Turn Debate Doesn’t Help Humans Answer Hard Reading Comprehension Questions](https://arxiv.org/abs/2210.10860) 301 | 302 |
303 | 304 | ### Tool use and scaffolding 305 | 306 | **Tier 2** 307 | 308 | - [Measuring the impact of post-training enhancements](https://metr.github.io/autonomy-evals-guide/elicitation-gap/) 309 | - [WebGPT: Browser-assisted question-answering with human feedback](https://arxiv.org/abs/2112.09332) 310 | 311 | **Tier 3** 312 | 313 | - ✨ [Executable Code Actions Elicit Better LLM Agents](https://arxiv.org/abs/2402.01030) 314 | - ✨ [GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning](https://arxiv.org/abs/2507.19457) 315 | - ✨ [TextGrad: Automatic "Differentiation" via Text](https://arxiv.org/abs/2406.07496) 316 | - [AI capabilities can be significantly improved without expensive retraining](http://arxiv.org/abs/2312.07413) 317 | - [Automated Statistical Model Discovery with Language Models](http://arxiv.org/abs/2402.17879) 318 | 319 |
Tier 4+ 320 | 321 | - [DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines](http://arxiv.org/abs/2310.03714) 322 | - [Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolution](http://arxiv.org/abs/2309.16797) 323 | - [Self-Taught Optimizer (STOP): Recursively Self-Improving Code Generation](https://arxiv.org/abs/2310.02304v1) 324 | - [Voyager: An Open-Ended Embodied Agent with Large Language Models](http://arxiv.org/abs/2305.16291) 325 | - [ReGAL: Refactoring Programs to Discover Generalizable Abstractions](http://arxiv.org/abs/2401.16467) 326 | 327 |
328 | 329 | ### Honesty, factuality, and epistemics 330 | 331 | **Tier 2** 332 | 333 | - [Self-critiquing models for assisting human evaluators](https://arxiv.org/abs/2206.05802v2) 334 | 335 | **Tier 3** 336 | 337 | - [What Evidence Do Language Models Find Convincing?](http://arxiv.org/abs/2402.11782) 338 | - [How to Catch an AI Liar: Lie Detection in Black-Box LLMs by Asking Unrelated Questions](https://arxiv.org/abs/2309.15840) 339 | 340 |
Tier 4+ 341 | 342 | - [Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting](http://arxiv.org/abs/2305.04388) 343 | - [Long-form factuality in large language models](http://arxiv.org/abs/2403.18802) 344 | 345 |
346 | 347 | ## Applications 348 | 349 | ### Science 350 | 351 | **Tier 2** 352 | 353 | - ✨ [AlphaEvolve: A coding agent for scientific and algorithmic discovery](https://arxiv.org/abs/2506.13131) 354 | - ✨ [AlphaFold 3: Accurate structure prediction of biomolecular interactions](https://www.nature.com/articles/s41586-024-07487-w) 355 | 356 | **Tier 3** 357 | 358 | - ✨ [Towards an AI Co-Scientist](https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/towards-an-ai-co-scientist/Towards_an_AI_Co-Scientist.pdf) 359 | - [Can large language models provide useful feedback on research papers? A large-scale empirical analysis](http://arxiv.org/abs/2310.01783) 360 | - [Large Language Models Encode Clinical Knowledge](http://arxiv.org/abs/2212.13138) 361 | - [The Impact of Large Language Models on Scientific Discovery: a Preliminary Study using GPT-4](http://arxiv.org/abs/2311.07361) 362 | - [A Dataset of Information-Seeking Questions and Answers Anchored in Research Papers](https://arxiv.org/abs/2105.03011) 363 | 364 |
Tier 4+ 365 | 366 | - [Can Generalist Foundation Models Outcompete Special-Purpose Tuning? Case Study in Medicine](http://arxiv.org/abs/2311.16452) 367 | - [Nougat: Neural Optical Understanding for Academic Documents](http://arxiv.org/abs/2308.13418) 368 | - [Scim: Intelligent Skimming Support for Scientific Papers](http://arxiv.org/abs/2205.04561) 369 | - [SynerGPT: In-Context Learning for Personalized Drug Synergy Prediction and Drug Design](https://www.biorxiv.org/content/10.1101/2023.07.06.547759v1) 370 | - [Towards Accurate Differential Diagnosis with Large Language Models](http://arxiv.org/abs/2312.00164) 371 | - [Towards a Benchmark for Scientific Understanding in Humans and Machines](http://arxiv.org/abs/2304.10327) 372 | - [A Search Engine for Discovery of Scientific Challenges and Directions](https://arxiv.org/abs/2108.13751) 373 | - [A full systematic review was completed in 2 weeks using automation tools: a case study](https://pubmed.ncbi.nlm.nih.gov/32004673/) 374 | - [Fact or Fiction: Verifying Scientific Claims](https://arxiv.org/abs/2004.14974) 375 | - [Multi-XScience: A Large-scale Dataset for Extreme Multi-document Summarization of Scientific Articles](https://arxiv.org/abs/2010.14235) 376 | - [PEER: A Collaborative Language Model](https://arxiv.org/abs/2208.11663) 377 | - [PubMedQA: A Dataset for Biomedical Research Question Answering](https://arxiv.org/abs/1909.06146) 378 | - [SciCo: Hierarchical Cross-Document Coreference for Scientific Concepts](https://arxiv.org/abs/2104.08809) 379 | - [SciTail: A Textual Entailment Dataset from Science Question Answering](http://ai2-website.s3.amazonaws.com/team/ashishs/scitail-aaai2018.pdf) 380 | 381 |
382 | 383 | ### Forecasting 384 | 385 | **Tier 3** 386 | 387 | - ✨ [Consistency Checks for Language Model Forecasters](https://arxiv.org/abs/2412.18544) 388 | - ✨ [LLM Processes: Numerical Predictive Distributions Conditioned on Natural Language](https://arxiv.org/abs/2405.12856) 389 | - [AI-Augmented Predictions: LLM Assistants Improve Human Forecasting Accuracy](https://arxiv.org/abs/2402.07862v1) 390 | - [Approaching Human-Level Forecasting with Language Models](http://arxiv.org/abs/2402.18563) 391 | - [Forecasting Future World Events with Neural Networks](https://arxiv.org/abs/2206.15474) 392 | 393 |
Tier 4+ 394 | 395 | - [Are Transformers Effective for Time Series Forecasting?](https://arxiv.org/abs/2205.13504) 396 | 397 |
398 | 399 | ### Search and ranking 400 | 401 | **Tier 2** 402 | 403 | - [Learning Dense Representations of Phrases at Scale](https://arxiv.org/abs/2012.12624) 404 | - [Text and Code Embeddings by Contrastive Pre-Training](https://arxiv.org/abs/2201.10005) (OpenAI embeddings) 405 | 406 | **Tier 3** 407 | 408 | - [Large Language Models are Effective Text Rankers with Pairwise Ranking Prompting](http://arxiv.org/abs/2306.17563) 409 | - [Not All Vector Databases Are Made Equal](https://dmitry-kan.medium.com/milvus-pinecone-vespa-weaviate-vald-gsi-what-unites-these-buzz-words-and-what-makes-each-9c65a3bd0696) 410 | - [REALM: Retrieval-Augmented Language Model Pre-Training](https://arxiv.org/abs/2002.08909) 411 | - [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://arxiv.org/abs/2005.11401) 412 | - [Task-aware Retrieval with Instructions](https://arxiv.org/abs/2211.09260) 413 | 414 |
Tier 4+ 415 | 416 | - [RankZephyr: Effective and Robust Zero-Shot Listwise Reranking is a Breeze!](http://arxiv.org/abs/2312.02724) 417 | - [Some Common Mistakes In IR Evaluation, And How They Can Be Avoided](https://dl.acm.org/doi/10.1145/3190580.3190586) 418 | - [Boosting Search Engines with Interactive Agents](https://arxiv.org/abs/2109.00527) 419 | - [ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT](https://arxiv.org/abs/2004.12832) 420 | - [Moving Beyond Downstream Task Accuracy for Information Retrieval Benchmarking](https://arxiv.org/abs/2212.01340) 421 | - [UnifiedQA: Crossing Format Boundaries With a Single QA System](https://arxiv.org/abs/2005.00700) 422 | 423 |
424 | 425 | ## ML in practice 426 | 427 | ### Production deployment 428 | 429 | **Tier 1** 430 | 431 | - [Machine Learning in Python: Main developments and technology trends in data science, machine learning, and AI](https://arxiv.org/abs/2002.04803v2) 432 | - [Machine Learning: The High Interest Credit Card of Technical Debt](https://papers.nips.cc/paper/2015/file/86df7dcfd896fcaf2674f757a2463eba-Paper.pdf) 433 | 434 | **Tier 2** 435 | 436 | - [Designing Data-Intensive Applications](https://dataintensive.net/) 437 | - [A Recipe for Training Neural Networks](http://karpathy.github.io/2019/04/25/recipe/) (Karpathy) 438 | 439 | ### Benchmarks 440 | 441 | **Tier 2** 442 | 443 | - ✨ [GAIA: a benchmark for General AI Assistants](http://arxiv.org/abs/2311.12983) 444 | - [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](http://arxiv.org/abs/2311.12022) 445 | - [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://arxiv.org/abs/2310.06770v1) 446 | - [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://arxiv.org/abs/2109.07958) 447 | 448 | **Tier 3** 449 | 450 | - ✨ [RE-Bench: Evaluating Frontier AI R&D Capabilities](https://arxiv.org/abs/2411.15114) 451 | - ✨ [SimpleQA: Measuring Short-Form Factuality](https://openai.com/index/introducing-simpleqa/) 452 | - ✨ [ARC Prize 2024: Technical Report](https://arcprize.org/blog/oai-o3-pub-breakthrough) 453 | - ✨ [FrontierMath: A Benchmark for Evaluating Advanced Mathematical Reasoning in AI](https://arxiv.org/abs/2411.04872) 454 | - [Measuring Massive Multitask Language Understanding](https://arxiv.org/abs/2009.03300) 455 | 456 |
Tier 4+ 457 | 458 | - [FLEX: Unifying Evaluation for Few-Shot NLP](https://arxiv.org/abs/2107.07170) 459 | - [Holistic Evaluation of Language Models](https://arxiv.org/abs/2107.07170) (HELM) 460 | - [True Few-Shot Learning with Language Models](https://arxiv.org/abs/2105.11447) 461 | - [ConditionalQA: A Complex Reading Comprehension Dataset with Conditional Answers](https://arxiv.org/abs/2110.06884) 462 | - [Measuring Mathematical Problem Solving With the MATH Dataset](https://arxiv.org/abs/2103.03874) 463 | - [QuALITY: Question Answering with Long Input Texts, Yes!](https://arxiv.org/abs/2112.08608) 464 | - [SCROLLS: Standardized CompaRison Over Long Language Sequences](https://arxiv.org/abs/2201.03533) 465 | - [What Will it Take to Fix Benchmarking in Natural Language Understanding?](https://arxiv.org/abs/2104.02145) 466 | 467 |
468 | 469 | ### Datasets 470 | 471 | **Tier 2** 472 | 473 | - [Common Crawl](https://arxiv.org/abs/2105.02732) 474 | - [The Pile: An 800GB Dataset of Diverse Text for Language Modeling](https://arxiv.org/abs/2101.00027) 475 | 476 | **Tier 3** 477 | 478 | - ✨ [FineWeb: Decanting the Web for the Finest Text Data at Scale](https://huggingface.co/spaces/HuggingFaceFW/blogpost-fineweb-v1) 479 | - [Dialog Inpainting: Turning Documents into Dialogs](https://arxiv.org/abs/2205.09073) 480 | - [MS MARCO: A Human Generated MAchine Reading COmprehension Dataset](https://arxiv.org/abs/1611.09268) 481 | - [Microsoft Academic Graph](https://internal-journal.frontiersin.org/articles/10.3389/fdata.2019.00045/full) 482 | 483 | ## Advanced topics 484 | 485 | ### World models and causality 486 | 487 | **Tier 3** 488 | 489 | - [Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Task](http://arxiv.org/abs/2210.13382) 490 | - [From Word Models to World Models: Translating from Natural Language to the Probabilistic Language of Thought](http://arxiv.org/abs/2306.12672) 491 | - [Language Models Represent Space and Time](http://arxiv.org/abs/2310.02207) 492 | 493 |
Tier 4+ 494 | 495 | - [Amortizing intractable inference in large language models](http://arxiv.org/abs/2310.04363) 496 | - [CLADDER: Assessing Causal Reasoning in Language Models](http://zhijing-jin.com/files/papers/CLadder_2023.pdf) 497 | - [Causal Bayesian Optimization](https://proceedings.mlr.press/v108/aglietti20a.html) 498 | - [Causal Reasoning and Large Language Models: Opening a New Frontier for Causality](http://arxiv.org/abs/2305.00050) 499 | - [Generative Agents: Interactive Simulacra of Human Behavior](http://arxiv.org/abs/2304.03442) 500 | - [Passive learning of active causal strategies in agents and language models](http://arxiv.org/abs/2305.16183) 501 | 502 |
503 | 504 | ### Planning 505 | 506 |
Tier 4+ 507 | 508 | - [Beyond A\*: Better Planning with Transformers via Search Dynamics Bootstrapping](http://arxiv.org/abs/2402.14083) 509 | - [Cognitive Architectures for Language Agents](http://arxiv.org/abs/2309.02427) 510 | 511 |
512 | 513 | ### Uncertainty, calibration, and active learning 514 | 515 | **Tier 2** 516 | 517 | - [Experts Don't Cheat: Learning What You Don't Know By Predicting Pairs](http://arxiv.org/abs/2402.08733) 518 | - [A Simple Baseline for Bayesian Uncertainty in Deep Learning](https://arxiv.org/abs/1902.02476) 519 | - [Plex: Towards Reliability using Pretrained Large Model Extensions](https://arxiv.org/abs/2207.07411) 520 | 521 | **Tier 3** 522 | 523 | - ✨ [Textual Bayes: Quantifying Uncertainty in LLM-Based Systems](https://arxiv.org/abs/2506.10060) 524 | - [Active Preference Inference using Language Models and Probabilistic Reasoning](http://arxiv.org/abs/2312.12009) 525 | - [Eliciting Human Preferences with Language Models](http://arxiv.org/abs/2310.11589) 526 | - [Describing Differences between Text Distributions with Natural Language](https://arxiv.org/abs/2201.12323) 527 | - [Teaching Models to Express Their Uncertainty in Words](https://arxiv.org/abs/2205.14334) 528 | 529 |
Tier 4+ 530 | 531 | - [Active Learning by Acquiring Contrastive Examples](https://arxiv.org/abs/2109.03764) 532 | - [Doing Experiments and Revising Rules with Natural Language and Probabilistic Reasoning](http://arxiv.org/abs/2402.06025) 533 | - [STaR-GATE: Teaching Language Models to Ask Clarifying Questions](http://arxiv.org/abs/2403.19154) 534 | - [Active Testing: Sample-Efficient Model Evaluation](https://arxiv.org/abs/2103.05331) 535 | - [Uncertainty Estimation for Language Reward Models](https://arxiv.org/abs/2203.07472) 536 | 537 |
538 | 539 | ### Interpretability and model editing 540 | 541 | **Tier 2** 542 | 543 | - ✨ [Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet](https://transformer-circuits.pub/2024/scaling-monosemanticity/) 544 | - ✨ [Interpretability at Scale: Identifying Causal Mechanisms in Alpaca](http://arxiv.org/abs/2305.08809) 545 | - [Discovering Latent Knowledge in Language Models Without Supervision](https://arxiv.org/abs/2212.03827v1) 546 | 547 | **Tier 3** 548 | 549 | - ✨ [Scaling and Evaluating Sparse Autoencoders](https://cdn.openai.com/papers/sparse-autoencoders.pdf) 550 | - ✨ [Opening the AI black box: program synthesis via mechanistic interpretability](https://arxiv.org/abs/2402.05110v1) 551 | - [Mechanistically analyzing the effects of fine-tuning on procedurally defined tasks](http://arxiv.org/abs/2311.12786) 552 | - [Representation Engineering: A Top-Down Approach to AI Transparency](http://arxiv.org/abs/2310.01405) 553 | - [Studying Large Language Model Generalization with Influence Functions](http://arxiv.org/abs/2308.03296) 554 | 555 |
Tier 4+ 556 | 557 | - [Codebook Features: Sparse and Discrete Interpretability for Neural Networks](http://arxiv.org/abs/2310.17230) 558 | - [Eliciting Latent Predictions from Transformers with the Tuned Lens](http://arxiv.org/abs/2303.08112) 559 | - [How do Language Models Bind Entities in Context?](http://arxiv.org/abs/2310.17191) 560 | - [Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small](https://arxiv.org/abs/2211.00593) 561 | - [Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models](http://arxiv.org/abs/2403.19647) 562 | - [Uncovering mesa-optimization algorithms in Transformers](http://arxiv.org/abs/2309.05858) 563 | - [Fast Model Editing at Scale](https://arxiv.org/abs/2110.11309) 564 | - [Git Re-Basin: Merging Models modulo Permutation Symmetries](https://arxiv.org/abs/2209.04836) 565 | - [Locating and Editing Factual Associations in GPT](https://arxiv.org/abs/2202.05262) 566 | - [Mass-Editing Memory in a Transformer](https://arxiv.org/abs/2210.07229) 567 | 568 |
569 | 570 | ### Reinforcement learning 571 | 572 | **Tier 2** 573 | 574 | - ✨ [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://arxiv.org/abs/2402.03300) (GRPO) 575 | - [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](http://arxiv.org/abs/2305.18290) 576 | - [Reflexion: Language Agents with Verbal Reinforcement Learning](http://arxiv.org/abs/2303.11366) 577 | - [Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm](https://arxiv.org/abs/1712.01815) (AlphaZero) 578 | - [MuZero: Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model](https://arxiv.org/abs/1911.08265) 579 | 580 | **Tier 3** 581 | 582 | - [Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback](http://arxiv.org/abs/2307.15217) 583 | - [AlphaStar: mastering the real-time strategy game StarCraft II](https://deepmind.com/blog/article/alphastar-mastering-real-time-strategy-game-starcraft-ii) 584 | - [Decision Transformer](https://arxiv.org/abs/2106.01345) 585 | - [Mastering Atari Games with Limited Data](https://arxiv.org/abs/2111.00210) (EfficientZero) 586 | - [Mastering Stratego, the classic game of imperfect information](https://www.science.org/doi/10.1126/science.add4679) (DeepNash) 587 | 588 |
Tier 4+ 589 | 590 | - [AlphaStar Unplugged: Large-Scale Offline Reinforcement Learning](http://arxiv.org/abs/2308.03526) 591 | - [Bayesian Reinforcement Learning with Limited Cognitive Load](http://arxiv.org/abs/2305.03263) 592 | - [Contrastive Prefence Learning: Learning from Human Feedback without RL](http://arxiv.org/abs/2310.13639) 593 | - [Grandmaster-Level Chess Without Search](http://arxiv.org/abs/2402.04494) 594 | - [A data-driven approach for learning to control computers](https://arxiv.org/abs/2202.08137) 595 | - [Acquisition of Chess Knowledge in AlphaZero](https://arxiv.org/abs/2111.09259) 596 | - [Player of Games](https://arxiv.org/abs/2112.03178) 597 | - [Retrieval-Augmented Reinforcement Learning](https://arxiv.org/abs/2202.08417) 598 | 599 |
600 | 601 | ## The big picture 602 | 603 | ### AI scaling 604 | 605 | **Tier 1** 606 | 607 | - [Scaling Laws for Neural Language Models](https://arxiv.org/abs/2001.08361) 608 | - [Takeoff speeds](https://sideways-view.com/2018/02/24/takeoff-speeds/) 609 | - [The Bitter Lesson](http://www.incompleteideas.net/IncIdeas/BitterLesson.html) 610 | 611 | **Tier 2** 612 | 613 | - [AI and compute](https://openai.com/blog/ai-and-compute/) 614 | - [Scaling Laws for Transfer](https://arxiv.org/abs/2102.01293) 615 | - [Training Compute-Optimal Large Language Models](https://arxiv.org/abs/2203.15556) (Chinchilla) 616 | 617 | **Tier 3** 618 | 619 | - ✨ [Pre-training under Infinite Compute](https://arxiv.org/abs/2509.14786) 620 | - [Emergent Abilities of Large Language Models](https://arxiv.org/abs/2206.07682) 621 | - [Transcending Scaling Laws with 0.1% Extra Compute](https://arxiv.org/abs/2210.11399) (U-PaLM) 622 | 623 |
Tier 4+ 624 | 625 | - [Physics of Language Models: Part 3.3, Knowledge Capacity Scaling Laws](http://arxiv.org/abs/2404.05405) 626 | - [Power Law Trends in Speedrunning and Machine Learning](http://arxiv.org/abs/2304.10004) 627 | - [Scaling laws for single-agent reinforcement learning](http://arxiv.org/abs/2301.13442) 628 | - [Beyond neural scaling laws: beating power law scaling via data pruning](https://arxiv.org/abs/2206.14486) 629 | - [Emergent Abilities of Large Language Models](https://arxiv.org/abs/2206.07682) 630 | - [Scaling Scaling Laws with Board Games](https://arxiv.org/abs/2104.03113) 631 | 632 |
633 | 634 | ### AI safety 635 | 636 | **Tier 1** 637 | 638 | - [Three impacts of machine intelligence](https://www.effectivealtruism.org/articles/three-impacts-of-machine-intelligence-paul-christiano/) 639 | - [What failure looks like](https://www.alignmentforum.org/posts/HBxe6wdjxK239zajf/what-failure-looks-like) 640 | - [Without specific countermeasures, the easiest path to transformative AI likely leads to AI takeover](https://www.lesswrong.com/posts/pRkFkzwKZ2zfa3R6H/without-specific-countermeasures-the-easiest-path-to) 641 | 642 | **Tier 2** 643 | 644 | - [An Overview of Catastrophic AI Risks](http://arxiv.org/abs/2306.12001) 645 | - [Clarifying “What failure looks like” (part 1)](https://www.lesswrong.com/posts/v6Q7T335KCMxujhZu/clarifying-what-failure-looks-like-part-1) 646 | - [Deep RL from human preferences](https://openai.com/blog/deep-reinforcement-learning-from-human-preferences/) 647 | - [The alignment problem from a deep learning perspective](https://arxiv.org/abs/2209.00626) 648 | 649 | **Tier 3** 650 | 651 | - ✨ [Alignment Faking in Large Language Models](https://arxiv.org/abs/2412.14093) 652 | - ✨ [Constitutional Classifiers: Defending against Universal Jailbreaks](https://arxiv.org/abs/2501.18837) 653 | - ✨ [Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs](https://arxiv.org/abs/2502.17424) 654 | - ✨ [Gradual Disempowerment: Systemic Existential Risks from Incremental AI Development](https://arxiv.org/abs/2501.16946) 655 | - [Scheming AIs: Will AIs fake alignment during training in order to get power?](http://arxiv.org/abs/2311.08379) 656 | 657 |
Tier 4+ 658 | 659 | - [Towards a Law of Iterated Expectations for Heuristic Estimators](https://arxiv.org/abs/2410.01290) 660 | - [Measuring Progress on Scalable Oversight for Large Language Models](https://arxiv.org/abs/2211.03540) 661 | - [Scalable agent alignment via reward modelling](https://arxiv.org/abs/1811.07871) 662 | - [AI Deception: A Survey of Examples, Risks, and Potential Solutions](http://arxiv.org/abs/2308.14752) 663 | - [Benchmarks for Detecting Measurement Tampering](http://arxiv.org/abs/2308.15605) 664 | - [Chess as a Testing Grounds for the Oracle Approach to AI Safety](http://arxiv.org/abs/2010.02911) 665 | - [Close the Gates to an Inhuman Future: How and why we should choose to not develop superhuman general-purpose artificial intelligence](https://papers.ssrn.com/abstract=4608505) 666 | - [Model evaluation for extreme risks](http://arxiv.org/abs/2305.15324) 667 | - [Responsible Reporting for Frontier AI Development](http://arxiv.org/abs/2404.02675) 668 | - [Safety Cases: How to Justify the Safety of Advanced AI Systems](http://arxiv.org/abs/2403.10462) 669 | - [Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training](http://arxiv.org/abs/2401.05566) 670 | - [Technical Report: Large Language Models can Strategically Deceive their Users when Put Under Pressure](http://arxiv.org/abs/2311.07590) 671 | - [Tensor Trust: Interpretable Prompt Injection Attacks from an Online Game](http://arxiv.org/abs/2311.01011) 672 | - [Tools for Verifying Neural Models' Training Data](http://arxiv.org/abs/2307.00682) 673 | - [Towards a Cautious Scientist AI with Convergent Safety Bounds](https://yoshuabengio.org/2024/02/26/towards-a-cautious-scientist-ai-with-convergent-safety-bounds/) 674 | - [Alignment of Language Agents](https://arxiv.org/abs/2103.14659) 675 | - [Eliciting Latent Knowledge](https://docs.google.com/document/d/1WwsnJQstPq91_Yh-Ch2XRL8H_EpsnjrC1dwZXR37PC8/edit?usp=sharing) 676 | - [Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned](https://arxiv.org/abs/2209.07858) 677 | - [Red Teaming Language Models with Language Models](https://storage.googleapis.com/deepmind-media/Red%20Teaming/Red%20Teaming.pdf) 678 | - [Risks from Learned Optimization in Advanced Machine Learning Systems](https://arxiv.org/abs/1906.01820) 679 | - [Unsolved Problems in ML Safety](https://arxiv.org/abs/2109.13916) 680 | 681 |
682 | 683 | ### Economic and social impacts 684 | 685 | **Tier 2** 686 | 687 | - ✨ [AI 2027](https://ai-2027.com/) 688 | - ✨ [Situational Awareness](https://situational-awareness.ai/) (Aschenbrenner) 689 | 690 | **Tier 3** 691 | 692 | - [Explosive growth from AI automation: A review of the arguments](http://arxiv.org/abs/2309.11690) 693 | - [Language Models Can Reduce Asymmetry in Information Markets](http://arxiv.org/abs/2403.14443) 694 | 695 |
Tier 4+ 696 | 697 | - [Bridging the Human-AI Knowledge Gap: Concept Discovery and Transfer in AlphaZero](http://arxiv.org/abs/2310.16410) 698 | - [Foundation Models and Fair Use](https://arxiv.org/abs/2303.15715v1) 699 | - [GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models](http://arxiv.org/abs/2303.10130) 700 | - [Levels of AGI: Operationalizing Progress on the Path to AGI](http://arxiv.org/abs/2311.02462) 701 | - [Opportunities and Risks of LLMs for Scalable Deliberation with Polis](http://arxiv.org/abs/2306.11932) 702 | - [On the Opportunities and Risks of Foundation Models](https://arxiv.org/abs/2108.07258) 703 | 704 |
705 | 706 | ### Philosophy 707 | 708 | **Tier 2** 709 | 710 | - [Meaning without reference in large language models](https://arxiv.org/abs/2208.02957) 711 | 712 |
Tier 4+ 713 | 714 | - [Consciousness in Artificial Intelligence: Insights from the Science of Consciousness](http://arxiv.org/abs/2308.08708) 715 | - [Philosophers Ought to Develop, Theorize About, and Use Philosophically Relevant AI](https://philarchive.org/archive/CLAPOT-16) 716 | - [Towards Evaluating AI Systems for Moral Status Using Self-Reports](http://arxiv.org/abs/2311.08576) 717 | 718 |
719 | 720 | ## Maintainer 721 | 722 | [andreas@elicit.com](mailto:andreas@elicit.com) 723 | --------------------------------------------------------------------------------