Are Latent Reasoning Models Easily Interpretable? Checkpoints for Dilgren & Wiegreffe (2026), arxiv.org/abs/2604.04902: No-CoT/CoT/Coconut/CODI + multi-mode, GPT-2 & Llama-3.2-1B. connordilgren/gpt2-gsm8k-no-cot Updated Jul 2 connordilgren/gpt2-gsm8k-cot Updated Jul 2 connordilgren/gpt2-gsm8k-coconut Updated Jul 2 connordilgren/gpt2-gsm8k-multimode-coconut Updated Jul 2
Are Latent Reasoning Models Easily Interpretable? Checkpoints for Dilgren & Wiegreffe (2026), arxiv.org/abs/2604.04902: No-CoT/CoT/Coconut/CODI + multi-mode, GPT-2 & Llama-3.2-1B. connordilgren/gpt2-gsm8k-no-cot Updated Jul 2 connordilgren/gpt2-gsm8k-cot Updated Jul 2 connordilgren/gpt2-gsm8k-coconut Updated Jul 2 connordilgren/gpt2-gsm8k-multimode-coconut Updated Jul 2
