
Nowait Reasoning Optimizer
This implements the NOWAIT technique from a 2025 paper that suppresses self-reflection tokens during inference to cut reasoning token usage by 27-51% without hurting accuracy. It works by blocking words like "wait," "hmm," and "alternatively" during generation, which forces models like QwQ and DeepSeek-R1 into more direct reasoning paths. Works great on RL-based models but can degrade performance on distilled ones since they rely heavily on the CoT structure from training. The implementation is straightforward, just a logit processor you drop into your generation call. If you're running reasoning models in production and token costs or latency matter, this is a clean win with minimal integration work.
- skills.sh installs
- 438
- GitHub stars
- 33.2K
- last updated
- Aug 21, 2026
Install
$ npx -y skills add davila7/claude-code-templates --skill nowait-reasoning-optimizer --agent claude-codeInstalls into .claude/skills of the current project.
Files
This implements the NOWAIT technique from a 2025 paper that suppresses self-reflection tokens during inference to cut reasoning token usage by 27-51% without hurting accuracy. It works by blocking words like "wait," "hmm," and "alternatively" during generation, which forces models like QwQ and DeepSeek-R1 into more direct reasoning paths. Works great on RL-based models but can degrade performance on distilled ones since they rely heavily on the CoT structure from training. The implementation is straightforward, just a logit processor you drop into your generation call. If you're running reasoning models in production and token costs or latency matter, this is a clean win with minimal integration work.
Browse the repository on GitHubSponsors
Details
- Repository
- github.com/davila7/claude-code-templates
- Category
- Code Review & Quality
- First seen
- Jun 3, 2026
More from davila7/claude-code-templates
All 618- Code Review Checklist437
- Prompt Caching437
- Commit435
- Vercel Deployment433
- C4 Architecture430
- Information Security Manager Iso27001428
- Mlflow427
- Statsmodels426
- Prompt Library423
- Chroma421
- Telegram Mini App420
- Pricing Strategy419
- Seo419
- Evaluating Llms Harness418
- Ml Paper Writing416
- Openalex Database416
- Workflow Automation416
- Backend To Frontend Handoff Docs413
- Shopify Development411
- Nextjs Supabase Auth410
- Clinicaltrials Database409
- Writing Plans408
- Geopandas407
- Bash Linux405