Official code for the paper: "Simple LLM Baselines are Competitive for Model Diffing"
-
Updated
Feb 13, 2026 - Python
#
model-diffing
Here are 10 public repositories matching this topic...
ActDiff: diagnose, repair, and prevent persistent domain priors in narrowly finetuned vision-language models.
-
Updated
Aug 21, 2026 - Python
Code and artifacts for The Convergence Gap: when instruction-tuned models settle on next-token predictions.
language-models reproducibility early-exit large-language-models instruction-tuning mechanistic-interpretability llm-interpretability model-diffing tuned-lens convergence-gap
-
Updated
Aug 11, 2026 - Python
Supplementary code & results for "Variant-specific crosscoder features are seed-stable but not detectably task-causal in a GRPO-LoRA math setting" (ICML 2026 Mech Interp Workshop, Spotlight)
reinforcement-learning sparse-autoencoders interpretability mechanistic-interpretability crosscoders model-diffing
-
Updated
Jun 11, 2026 - Jupyter Notebook
Mechanistic study of contextual-integrity post-training in Qwen2.5-7B, testing whether improved privacy behavior comes from new mechanisms or better use of machinery already present in the base model.
privacy language-model ai-safety interpretability qwen-7b llm-interpretability model-diffing causal-interventions contextual-integrity activation-patch
-
Updated
Sep 22, 2026 - Python
Behavioral auditing of finetuned language models using blinded policy recovery, model diffing, and causal activation interventions.
alignment ai-safety fine-tuning mechanistic-interpretability llm-evaluation model-auditing model-diffing causal-interventions
-
Updated
Sep 21, 2026 - Python
Sealed, preregistered benchmark for black-box model-diffing agents: five LoRA finetunes of Qwen3.5-9B (one null, three planted behaviours, one dropped backdoor), audited blind by Neel Nanda's diffing-agent recipe and four cheaper conditions. The recipe fails by not asking, and the auditor itself is a failure mode. MATS 12 application.
-
Updated
Sep 5, 2026 - Python
CDD Diffing Technique [Contrastive Decoding Diffing]
diffing mechanistic-models mechanistic-empirical mechanistic-inference rlhf mechanistic-interpretability mechanistic-modeling diffing-algorithm rlhf-v model-differencing mechanistic-intepretability rlhf-alternative rlhf-reward-model model-diffing mechanistic-validity rlhf-ppo
-
Updated
Sep 25, 2026 - Python
Code and artifacts for Same Targets, Different Computation: how post-training divides work across model layers.
language-models reproducibility fine-tuning large-language-models instruction-tuning mechanistic-interpretability activation-patching llm-interpretability model-diffing cross-patching
-
Updated
Aug 11, 2026 - Python
Research workspace for model diffing between pretrained and post-trained language models.
transcoder llama sae post-training instruction-tuning mechanistic-interpretability qwen deepseek olmo llm-interpretability gemma3 model-diffing tuned-lens cross-patching
-
Updated
Aug 11, 2026 - Python
Add this topic to your repo
To associate your repository with the model-diffing topic, visit your repo's landing page and select "manage topics."