Thomas Kwa
3 papers · 2024–2024 · 1 venue · across top CS/AI venues
Authors are matched by name, so researchers who share a name ("Yang Liu", "Wei Wang") appear as a single profile and their paper counts, venues and achievements are combined. How identity is handled.
Achievements
Jump to papers ↓
🌉
Interdisciplinary Bridge · Bronze
🗺️
Taxonomy Completionist
(14)
🐣
Hot Topic Early Bird
Venues
Top co-authors
Research topics
Artificial Intelligence
(3)
Deep Learning
(3)
Core AI
(3)
Machine Learning
(2)
Architectures
(2)
Learning Types · Machine Learning
(2)
Learning Types · Deep Learning
(2)
Optimization & Theory · Deep Learning
(2)
Reinforcement Learning
(1)
Application Areas
(1)
Methods
(1)
Optimization & Theory · Machine Learning
(1)
Keywords
mechanistic interpretability
(2)
neural network
(2)
neural network verification
(2)
policy optimization
(1)
kl divergence
(1)
reinforcement learning from human feedback
(1)
formal verification
(1)
heavy-tailed distribution
(1)
reward misspecification
(1)
reward hacking
(1)
formal guarantee
(1)
accuracy lower bound
(1)
proof transferability
(1)
causal model
(1)
circuit discovery
(1)
interchange intervention training
(1)
performance bound
(1)
transformer model
(1)
transformer architecture
(1)
accuracy bound
(1)
Papers
Catastrophic Goodhart: regularizing RLHF with KL divergence does not mitigate heavy-tailed reward misspecification
NeurIPS 2024
Compact Proofs of Model Performance via Mechanistic Interpretability
NeurIPS 2024
InterpBench: Semi-Synthetic Transformers for Evaluating Mechanistic Interpretability Techniques
NeurIPS 2024