AI
Grab a Coffee: Future-Aware Guidance for Discrete Diffusion with Compiled Objectives
Publication: cs.AI updates on arXiv.org
Author: Hua (Edward), Xu, Dongxin Li, Gwen Yidou-Weng, Guy Van den Broeck, Wei Wang, Anji Liu
Source published: 30/09/2026, 04:00:00 UTC
Briefing created: 30/09/2026, 04:20:14 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 30/09/2026, 04:03:25 UTC.
Briefings
COFFEE is a new framework for discrete diffusion models that separates sequence-level guidance from token-by-token computation. It uses a target-free carrier to absorb marginal token distributions and a compiled finite-state model to record how combinations affect sequence preferences, enabling joint conditioning without enumerating all completions. Evaluated on symbolic, language, and biological benchmarks, COFFEE achieves strong control with task-dependent quality and diversity trade-offs, demonstrating the potential of neural-symbolic methods in diffusion guidance.
Key points
- COFFEE separates sequence-level objectives from token-by-token computation to avoid exponential growth in guidance computation.
- The framework uses a target-free carrier to absorb marginal token distributions and a compiled finite-state model to record sequence-level preferences.
Context and limitations
Evaluation was conducted across multiple symbolic, language, and biological benchmarks, but specific results for each domain are not detailed in the evidence.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 30/09/2026, 04:20:14 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeidb5k2pzohkhhflq5b65fwlrpvtzjt3uvsw7pvphznagc7t5ipeia