Ms. Pac-Man RL neuron interpretability - a BurnyCoder Collection

BurnyCoder 's Collections

Ms. Pac-Man RL neuron interpretability

auto-eureka-scientist

eureka-digits-luna

eureka-digits-haiku

Geometry of Mathematics in LLMs

Accurate and simple interpretability explanations

Ant locomotion steering evidence

Mamba neuron interpretability

End-to-End BSFs on DINOv3 - full

End-to-End BSFs on DINOv3 - smoke

GPhyT Attention Residuals

Looped-LM neurons

Counterfactual concept swaps in CNNs

MiniStrike staged shooter self-play

Qwen3-0.6B GRAFT with LLM-as-a-Judge and Probe Monitors

Visualizing grokking modular addition

Oscillatory flow matching — local exploratory study

4D Snake RL: all evaluated networks

Atemokoloporos Qwen3.8-27B LoRA runs

Atemokoloporos Qwen3.5-0.8B retained checkpoints

Block-Sparse Featurizers on DINOv3 Rabbits

Ms. Pac-Man RL neuron interpretability

updated about 13 hours ago

Neuron-level interpretability and steering of reward-trained (no imitation) Ms. Pac-Man RL agents.

edbeeching/atari_2B_atari_mspacman_1111

Reinforcement Learning • Updated Apr 16, 2024 • 12 • 1

Note Pretrained APPO agent analysed in the study
edbeeching/atari_2B_atari_mspacman_2222

Reinforcement Learning • Updated Apr 16, 2024 • 13

Note Pretrained APPO agent analysed in the study
edbeeching/atari_2B_atari_mspacman_3333

Reinforcement Learning • Updated Apr 16, 2024 • 8

Note Pretrained APPO agent analysed in the study
BurnyCoder/mspacman-appo-neuron-atlas

Updated about 12 hours ago • 50

Note Derived results, unit atlas and analysis arrays. Code, paper and preregistration: https://github.com/BurnyCoder/pacman-reinforcement-learning-neuron-mechanistic-interpretability