Plan Canvas: Fixed Reasoning Regions for Continuous Language Flows
cs.CL
Submitted: 2026-10-05
Updated: 2026-10-05
License: http://creativecommons.org/licenses/by/4.0/
The gist: Continuous language flows generate text by denoising all positions of a target canvas together.
Abstract
Continuous language flows generate text by denoising all positions of a target canvas together. The natural way to add reasoning to such a model is to write a trace ahead of the answer, but the trace length changes from question to question. The answer start is therefore unknown during denoising, and the model has to decide the trace length, the place of every trace token, and the answer at the same time. We propose Plan Canvas to fix the boundary between the trace and the answer. A plan region of fixed capacity holds a compact trace, supervised padding fills its unused positions, and the answer starts at a fixed position. The fixed regions also allow separate denoising clocks for the plan and for the answer. With the trace text, backbone, and canvas length of the free-trace baseline held fixed, Plan Canvas improves accuracy on ProsQA and on Deep ProsQA, a graph benchmark with longer proofs. On Deep ProsQA, accuracy rises from 73.0% to 87.0%, the share of questions answered with a valid path rises from 30.8% to 59.1%, and the gain is largest on the longest proofs.
Sources
- Analog Bits: Generating Discrete Data using Diffusion Models with Self-Conditioning
- Compressed Chain of Thought: Efficient Reasoning Through Dense Representations
- Implicit Chain of Thought Reasoning via Knowledge Distillation
- Continuous diffusion for categorical data
- DiffuSeq: Sequence to Sequence Text Generation with Diffusion Models
- Training Large Language Models to Reason in a Continuous Latent Space
- Classifier-Free Diffusion Guidance
- ELF: Embedded Language Flows
- From Interface to Inference: Eliciting Any-Order Inference from Any-Order Models
- Flow Matching for Generative Modeling
- Muon is Scalable for LLM Training
- Let's Think Dot by Dot: Hidden Computation in Transformer Language Models
- Self-conditioned Embedding Diffusion for Text Generation
- Latent Chain-of-Thought as Planning: Decoupling Reasoning from Verbalization
- Self-Consistency Improves Chain of Thought Reasoning in Language Models
- Chain of Draft: Thinking Faster by Writing Less
- Answer First, Reason Later: When Commitment Order Costs Accuracy in Diffusion Language Models
- Coevolutionary Continuous Discrete Diffusion: Make Your Diffusion Language Model a Latent Reasoner
- Least-to-Most Prompting Enables Complex Reasoning in Large Language Models
Related papers
- Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
- Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
- Subliminal Steering: Stronger Encoding of Hidden Signals
- MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
- The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
- Untangling the Mechanisms of Misleading Context in Medical Question Answering