Where Should Diffusion Enter a Language Model? Geometry-Guided Hidden-State Replacement
For researchers building hybrid diffusion-transformer language models, this work provides a principled way to identify effective insertion points, though results are diagnostic and incremental.
DiHAL proposes a geometry-guided method to insert a diffusion bridge into a pretrained transformer by replacing lower layers, achieving improved hidden-state recovery over continuous diffusion baselines on 8B-scale models under a fixed training budget.
Continuous diffusion language models lag behind autoregressive transformers, partly because diffusion is applied in spaces poorly suited to language denoising and token recovery. We propose DiHAL, a geometry-guided diffusion-transformer hybrid that asks where diffusion should enter a pretrained transformer. DiHAL scores layers with geometry-based proxies, selects a diffusion-friendly hidden-state interface, and replaces the lower transformer prefix with a diffusion bridge while retaining the upper layers and original LM head. By reconstructing the selected-layer hidden state rather than tokens, DiHAL avoids direct continuous-to-discrete recovery. Experiments on 8B-scale backbones show that the geometry score predicts effective shallow insertion layers under a fixed bridge-training protocol and that hidden-state recovery improves over continuous diffusion baselines in a diagnostic comparison matching the diffusion/recovery training budget. These results suggest that hidden-state geometry helps identify where diffusion-based replacement is feasible inside pretrained language models.