Blackboard intelligence improves complex problem solving beyond autoregressive methods
Blackboard Intelligence Can Surpass Autoregressive on Globally Constrained Problems
Machine LearningComputation and Language
Summary
Large language models usually predict the next word one at a time, which can be tricky for problems that need global coordination. The authors show that using a different approach—where the model works on a shared workspace and revises its answers—helps solve these complex tasks better. They built their method using diffusion models that predict parts of solutions flexibly. Experiments show their method beats traditional techniques on puzzles in logic, scheduling nurses, and job-shop planning.
What this means in practice
- •For operations planners: Generate better schedules meeting complex constraints for workforce and manufacturing using blackboard-guided search.
- •For software engineers: Improve AI-based problem-solving modules handling globally constrained tasks by integrating blackboard inference with existing language models.
Authors
Woosang Jeon, Jaeyeon Kim, Sham Kakade, Yilun Du, Amrit Singh Bedi, Arun Kumar Chithanar, Chul Lee, Taehyeong Kim, Sitan Chen
Abstract
Next-token prediction has driven remarkable progress in large language models, yet a growing body of evidence suggests that they can struggle on problems governed by complex global constraints. In this work, we focus on this regime and ask whether some of these limitations arise from the inference interface induced by next-token prediction itself. We study this question through blackboard intelligence: an inference-time perspective in which a model works on a fixed, revisable canvas and searches over candidate solution states rather than committing to a causal, left-to-right trajectory. We instantiate this idea with diffusion language models, whose any-order prediction interface naturally exposes predictions over partially filled solution states. Our key observation is that mean confidence, a simple model-internal quantity available from the standard masked diffusion objective, provides a useful proxy for global coherence and can guide inference-time search and revision. Empirically, across ZebraLogic, Nurse Rostering, and Job-Shop Scheduling, Blackboard consistently improves inference while holding the fine-tuned LLaDA-8B-Instruct checkpoint fixed and substantially outperforms same-scale autoregressive baselines, reaching 90.4% accuracy on ZebraLogic-Hard, 76.4% exact feasibility on Nurse Rostering, and 80.2% optimality on JSSP. Stronger autoregressive search and refinement also fail to close the gap on ZebraLogic-Hard, while Blackboard surpasses tested frontier LLMs there and on JSSP despite their substantially greater scale and strong test-time reasoning. We open-source our codebase at https://github.com/jwoosang1/blackboard-intelligence.