Reference library
Improved Large Language Diffusion Models
Paper summary
iLLaDA scales masked-diffusion language-model training with bidirectional attention and a consistent objective through supervised fine-tuning. Variable-length generation and evaluation changes accompany improved results over the earlier diffusion model.
Relevance to the project
An alternative to autoregressive language modeling for optional model work. Useful when separating architecture effects from data scale, training stages, and evaluation changes.
Bidirectional text generation does not satisfy the engine's causal-history contract automatically.
Source record
- Reference ID
- illada
- Archived version
- 2606.25331v1
- Summary provenance
- reference-library/papers/illada/metadata.json · summary
- Review status
- initial orientation summary; not a critical full-paper review
- Reading scope
- Selected abstract/opening and available closing discussion/conclusion passages in the preserved text extract; broader project notes provide context. No reproduction of experiments, exhaustive methods review, or equation/figure audit.