AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

The Hidden Cost of Structured Generation in LLMs: Draft-Conditioned Constrained Decoding

Avinash Reddy, Thayne T. Walker, Jaime S. Ide, Amrit Singh Bedi

ICML 2026 · Large language models › Language, safety and interpretability · approach: Parallel and early-exit decoding · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add draft conditioning before constrained decoding (DCCD vs CD)Llama-3.2-1B constrained decoding · GSM8Kstrict structured accuracy: 15.24 -> 39.04Improved
add draft conditioning (DCCD vs CD)Qwen2.5-1.5B constrained decoding · MATH500strict structured accuracy: 15.00 -> 38.20Improved
add draft conditioning (DCCD vs CD)Qwen2.5-14B constrained decoding · GSM8Kstrict structured accuracy: 86.43 -> 95.15Improved
add draft conditioning (DCCD vs CD)Llama-3.2-1B constrained decoding · GSM Symbolicstrict structured accuracy: 0.00 -> 9.00Improved
add draft conditioning (DCCD vs CD)Qwen2.5-7B constrained decoding · FOLIOstrict structured accuracy: 19.70 -> 31.53Improved
average over all benchmarks, DCCD vs CD at 1BLlama-3.2-1B · average of four benchmarksstrict accuracy: 10.2% -> 20.9%Improved
add decode-time grammar masking on top of structural prompt (ablation)Qwen2.5-1.5B · MATH500answer accuracy: 41.4 -> 15.0Hurt
add decode-time grammar masking on top of structural prompt (ablation)Qwen2.5-14B · MATH500answer accuracy: 70.8 -> 47.6Hurt
mention the structural schema in the prompt (no CD) (ablation)Qwen2.5-14B · MATH500answer accuracy: 76.0 -> 70.8Hurt
mention the structural schema in the prompt (no CD) (ablation)Qwen2.5-7B · MATH500answer accuracy: 71.4 -> 71.2No clear effect
compose 7B draft model plus 1.5B constrainer vs single 14B CDDCCD composition · GSM8Kaccuracy per billion parameters: 6.1 -> 10.7Improved
compose 1.5B+1.5B DCCD vs 8B CDDCCD composition · MATH500accuracy per billion parameters: 3.6 -> 12.7Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy