Research map / Large language models / Instruction and fine-tuning
Cubing for Tuning
Haoze Wu, Clark Barrett, Nina Narodytska
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 |
|---|---|---|---|
| add online strategy tuning (TACO) to cube-and-conquer | CNC with Kissat · CnC and cruxmiter SAT benchmarks | total conflicts and runtime: conflicts reduced on all benchmarks; runtime reduced over 15% on 12 of 24 | Improved |
| add TACO to CNC | CNC with Kissat · SC-hard (grs 192 256) | solving time (s): 1390 -> 1963 (conflicts reduced) | Hurt |
| add TACO to CNC | CNC with Kissat · SC-hard (unsolved SAT competition instances) | benchmarks solved: solves php15 mixed 15 that CNC cannot; conflicts consistently reduced | Improved |
| add TACO to CNC | CNC with Marabou · NAP (235 NNV queries) | benchmarks solved: 163 -> 216 | Improved |
| add TACO to CNC | CNC with Marabou · AltLoop (259 NNV queries) | total time (s): 93764 -> 66532 | Improved |
| use learned new strategy (less input splitting) on benchmarks where one was found | CNC+TACO with Marabou · AltLoop (53 benchmarks) | total time (s): 46761 -> 26161 (44%) | Improved |
| learn bump=0 and tumble=0 (static branching order) | CNC+TACO with Kissat · eq.atree.braun.13 | runtime (s): 243 -> 139 | Improved |
Closest papers by meaning
- Learning How to Cube (NeurIPS 2026 · similarity 0.87)
- Learning Iterative Reasoning through Energy Diffusion (ICML 2024 · similarity 0.84)
- DART-Math: Difficulty-Aware Rejection Tuning for Mathematical Problem-Solving (NeurIPS 2024 · similarity 0.84)
- SolverLLM: Leveraging Test-Time Scaling for Optimization Problem via LLM-Guided Search (NeurIPS 2025 · similarity 0.83)
- Hint Tuning: Less Data Makes Better Reasoners (NeurIPS 2026 · similarity 0.83)
- Teaching Transformers to Solve Combinatorial Problems through Efficient Trial & Error (NeurIPS 2025 · similarity 0.83)
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.
