Balance-Aware Pivot Selection for Parallel Diffusion LLM Decoding
Diffusion language models enable parallel token generation, offering a promising alternative to sequential autoregressive decoding.
ProofPaper ↗
Key points
- Existing acceleration methods use predictive confidence to enable parallel token commitment while preserving generation quality.
- However, confidence-based selection prioritizes reliable token predictions without explicitly considering how token positions shape the remaining decoding schedule.
- Motivated by this limitation, we propose Balance-Aware Pivot Decoding (BPD), which favors central pivots within contiguous masked intervals and decodes the selected positions in parallel.
- Balanced pivots yield logarithmic decoding depth in the interval length.
Sources (1)
- [1]Balance-Aware Pivot Selection for Parallel Diffusion LLM DecodingarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 09:01 AM
Diffusion language models enable parallel token generation, offering a promising alternative to sequential autoregressive decoding.
Existing acceleration methods use predictive confidence to enable parallel token commitment while preserving generation quality.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 19, 2026Keep the Lineages Alive: Diversity-Preserving SMC for Diffusion LLM Decoding
- Oct 19, 2026Bayesian Model Discovery with Language Models: Closing the Loop between Hypotheses and Experiments
- Oct 18, 2026Enhanced Masked Diffusion Unlearning
- Oct 17, 2026Bellman Language Models: Solving the Recurring Decisions of LLM Agents
- Oct 17, 2026PV-Decoding: Progressive Verification Decoding with Decode Gating for Diffusion Language Models
- Oct 16, 2026[Research] 2,788 tokens/s on a single RTX 5090 with a diffusion LLM — Introducing ORL (fully open source)