- 1 Early Indicators of Reward Hacking via Reasoning Interpolation ↗
- 2 Reward Hacking Resarch Update ↗
- 3 Pretraining Data Filtering for Open-Weight AI Safety ↗
- 4 Attention Probes ↗
- 5 Research Update: Applications of Local Volume Measurement ↗
- 6 Studying inductive biases of random networks via local volumes ↗
- 7 The Common Pile v0.1 ↗
- 8 Product Key Memory Sparse Coders ↗
- 9 SAEs trained on the same data don’t learn the same features ↗
- 10 Partially rewriting an LLM in natural language ↗
- 11 Third-party evaluation to identify risks in LLMs’ training data ↗
- 12 Mechanistic Anomaly Detection Research Update 2 ↗
- 13 RLHF and RLAIF in GPT-NeoX ↗
- 14 The Practitioner's Guide to the Maximal Update Parameterization ↗
- 15 Mechanistic Anomaly Detection Research Update ↗
- 16 Open Source Automated Interpretability for Sparse Autoencoder Features ↗
- 17 Experiments in Weak-to-Strong Generalization ↗
- 18 Free Form Least-Squares Concept Erasure Without Oracle Concept Labels ↗
- 19 VINC-S: Closed-form Optionally-supervised Knowledge Elicitation with Paraphrase Invariance ↗
- 20 Pile-T5 ↗
- 21 Yi-34B, Llama 2, and common practices in LLM training: a fact check of the New York Times ↗
- 22 The Foundation Model Development Cheatsheet ↗
- 23 Least-Squares Concept Erasure with Oracle Concept Labels ↗
- 24 Diff-in-Means Concept Editing is Worst-Case Optimal ↗
- 25 The third New England RLHF Hackers Hackathon ↗
- 26 Extending the RoPE ↗
- 27 How the Foundation Model Transparency Index Distorts Transparency ↗
- 28 Llemma: An Open Language Model For Mathematics ↗
- 29 The second New England RLHF Hackers Hackathon ↗
- 30 Contributor Spotlight: Mohammad Aflah Khan ↗
- 31 The first New England RLHF Hackers Hackathon ↗
- 32 EleutherAI's Thoughts on the EU AI Act ↗
- 33 Minetester: A fully open RL environment built on Minetest ↗
- 34 🐶Safetensors audited as really safe and becoming the default ↗
- 35 Alignment Research @ EleutherAI ↗
- 36 Transformer Math 101 ↗
- 37 Exploratory Analysis of TRLX RLHF Transformers with TransformerLens ↗
- 38 EleutherAI Second Retrospective: The long version ↗
- 39 The View from 30,000 Feet: Preface to the Second EleutherAI Retrospective ↗
- 40 Announcing GPT-NeoX-20B ↗
- 41 A Preliminary Exploration into Factored Cognition with Language Models ↗
- 42 Multiple Choice Normalization in LM Evaluation ↗
- 43 Downstream Evaluations of Rotary Position Embeddings ↗
- 44 What A Long, Strange Trip It's Been: EleutherAI One Year Retrospective ↗
- 45 Why Release a Large Language Model? ↗
- 46 On the Sizes of OpenAI API Models ↗
- 47 Evaluating Different Fewshot Description Prompts on GPT-3 ↗
- 48 Finetuning Models on Downstream Tasks ↗
- 49 Activation Function Ablation ↗
- 50 Rotary Embeddings: A Relative Revolution ↗