Reading List

2024.09

• Evaluating Open-Source Sparse Autoencoders on Disentangling Factual Knowledge in GPT-2 Small
• Failures in Kindness
• Fun Theory
• Chai-1
• DataGemma
• LLaMa-Omni
• Agent Workflow Memory
• Learning to Reason with LLMs
• Can LLMs Generation Novel Research Ideas
• SimPO (Preference Optimization)
• Self-Contrastive Decoding with MoE
• Navigating the Safety Landscape: Measuring Risks in Finetuning Large Language Models
• Multi-Object Hallucination in Vision-Language Models

• XplainLLM: A Knowledge-Augmented Dataset for Reliable Grounded Explanations in LLMs

• Entity-Based Knowledge Conflicts in Question Answering
• DisentQA: Disentangling Parametric and Contextual Knowledge with Counterfactual Question Answering
• Editing Factual Knowledge in Language Models

• Adaptable and Interpretable Neural MemoryOver Symbolic Knowledge

• Uncovering Intermediate Variables in Transformers using Circuit Probing
• Transformer Circuit Faithfulness Metrics Are Not Robust
• Multilevel Interpretability Of Artificial Neural Networks: Leveraging Framework And Methods From Neuroscience
• Bi-Factorial Preference Optimization: Balancing Safety-Helpfulness in Language Models
• Capabilities of Gemini Models in Medicine, 2024
• Advancing Multimodal Medical Capabilities of Gemini, 2024

2024.08

• Enhancing Court View Generation with Knowledge Injection and Guidance
• Precedent-Enhanced Legal Judgment Prediction with LLM and Domain-Model Collaboration
• Discovering Bias in Latent Space: An Unsupervised Debiasing Approach
• Google: 6 incredible images of the human brain built with the help of Google’s AI
• 신경학 기초: 뉴런의 기능, 애니메이션
• Detoxifying Large Language Models via Knowledge Editing
• Towards Understanding Sycophancy in Language Models
• Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
• Epistemic Injustice in Generative AI
• A Mechanistic Understanding of Alignment Algorithms: A Case Study on DPO and Toxicity
• Recurrent Neural Networks Learn to Store and Generate Sequences using Non-Linear Representations
• Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models
• The Linear Representation Hypothesis and the Geometry of Large Language Models (ICML 2024)
• Function Vectors (ICLR 2024)
• Limitations on Formal Verification for AI Safety (Less Wrong)
• Provably safe systems: the only path to controllable AGI (Arxiv)
• You don’t know how bad most things are nor precisely how they’re bad. (Less Wrong)