LLM Digest

Story

arxiv_cs_lg · Nov 8, 2026 · paper

Source brief

D3Tree: Learning to Prune Block-Diffusion Draft Trees for Efficient Verification

arxiv.orgNov 8, 2026
original source linked

In brief

Block-diffusion drafters propose a whole block of tokens in one forward pass, and draft trees built from their position-wise marginals increase the expected accepted length of speculative decoding. Every tree node occupies a query row in target verification, so under concurrent s

Read the original at arxiv.org →Open in live feed Read that day’s brief

D3Tree: Learning to Prune Block-Diffusion Draft Trees for Efficient Verification

Earlier in this thread 4 items

Prune Gradually, Adapt Sparsely: Efficient Pruning of Large Language Models at Extreme Sparsity

Verify Smarter, Evolve Further: Efficient Harness Evolution through Behavior-Aware Verification

DARTree: Speculative Diffusion Decoding with Autoregressive Draft Trees

Coordinated Mixed-Precision KV Cache Quantization for Efficient LLM-based Multi-agent Systems