AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Mitigating Bias in Locally Constrained Decoding via Tractable Proposals

Meihua Dang, Linxin Song, Honghua Zhang, Jieyu Zhao, Guy Van den Broeck, Stefano Ermon

ICML 2026 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Sequential Monte Carlo for inference · 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
mask with global feasibility within token budget (GCD) instead of local masking (LCD)SMC proposal · xLAM JSON, k=4constraint satisfaction rate (%): 95.26 -> 100.0Improved
mask with global feasibility within token budget (GCD) instead of local masking (LCD)SMC proposal · xLAM (JSON and Python-like)accuracyImproved
add HMM probabilistic information (P-GCD) on top of GCDSMC proposal · xLAMconvergence speedImproved
use P-GCD (H=16384) instead of LCD proposalSMC on xLAM JSON · xLAM JSON, k=16accuracy (%): 69.7 -> 71.6Improved
use P-GCD with single particleSMC on xLAM JSON · xLAM JSON, k=1accuracy (%): 66.4 -> 66.5No clear effect
increase HMM hidden size from 1024 to 16384 (ablation)P-GCD · xLAM JSONaccuracy (%): k=2 68.6 -> 69.2; k=16 71.5 -> 71.6Improved
increase HMM hidden size to 16384 (ablation)P-GCD · xLAM latencyms/token: 5.26 -> 17.30Hurt
use GCD and P-GCD instead of LCD on SpiderSMC proposal · Spider, k=8execution accuracy (%): 67.2 -> 69.0 (GCD), 69.1 (P-GCD)Improved
use P-GCD proposal with a single particle on SpiderSMC proposal · Spider, k=1execution accuracy (%): 59.8 -> 56.2Hurt
add P-GCD potential to P-GCD proposal (ablation)SMC on CommonGen · CommonGen (GPT2-Large)BLEU-4Improved
use NFAs instead of DFAs for SQL constraintconstraint automaton · Spider SQL grammar, character levelnumber of states: 57,263 -> 7,999Improved
use tensorized GCD instead of unconstrained decodingdecoding · xLAM latencyms/token: 1.69 -> 1.86No clear effect

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