Research map / Learning theory and algorithms / Submodular and game algorithms
Greedy Flipping for Constrained Word Deletion
Jin-ge Yao, Xiaojun Wan
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace ILP solver with randomized constrained greedy flipping | LM+Disc objective · Written corpus | inference time per sentence (s): 1.72 -> 0.28 | Improved |
| replace ILP solver with randomized constrained greedy flipping | LM+Disc objective · Spoken corpus | inference time per sentence (s): 1.49 -> 0.25 | Improved |
| replace ILP solver with randomized constrained greedy flipping | LM+Disc objective · Written corpus | GR-F1 (%): 66.5 -> 66.4 | No clear effect |
| replace ILP solver with randomized constrained greedy flipping | LM+Disc objective · Spoken corpus | GR-F1 (%): 66.2 -> 66.2 | No clear effect |
| replace ILP solver with randomized constrained greedy flipping | LM objective · Written corpus | inference time per sentence (s): 1.23 -> 0.25 | Improved |
| replace ILP solver with randomized constrained greedy flipping | Disc objective · Spoken corpus | GR-F1 (%): 56.2 -> 55.9 | No clear effect |
| add discriminative scores to language model scores in the objective (ablation) | RCGF · Written corpus | GR-F1 (%): 64.5 -> 66.4 | Improved |
| add discriminative scores to language model scores in the objective (ablation) | ILP · Spoken corpus | GR-F1 (%): 61.5 -> 66.2 | Improved |
| use only discriminative scores instead of language model scores (ablation) | ILP · Written corpus | GR-F1 (%): 64.5 -> 62.0 | Hurt |
| add explicit syntactic constraints (ILP/RCGF) versus unconstrained baselines | RCGF(LM+Disc) · Written corpus | GR-F1 (%): 52.4 -> 66.4 | Improved |
| use RCGF instead of ILP in human evaluation | LM+Disc · Written corpus manual rating | grammaticality (1-5): 4.56 -> 4.56 | No clear effect |
| add explicit constraints versus McDonald06 in human evaluation | RCGF(LM+Disc) · Spoken corpus manual rating | grammaticality (1-5): 3.77 -> 4.38 | Improved |
Closest papers by meaning
- Grammar-Aligned Decoding (NeurIPS 2024 · similarity 0.82)
- Compressing LLMs: The Truth is Rarely Pure and Never Simple (ICLR 2024 · similarity 0.82)
- Constrained Entropic Unlearning: A Primal-Dual Framework for Large Language Models (NeurIPS 2025 · similarity 0.82)
- Fundamental Limits of Prompt Compression: A Rate-Distortion Framework for Black-Box Language Models (NeurIPS 2024 · similarity 0.82)
- Constrained Sampling for Language Models Should Be Easy: An MCMC Perspective (NeurIPS 2025 · similarity 0.82)
- A Partition Cover Approach to Tokenization (NeurIPS 2025 · similarity 0.82)
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.
