nowait-reasoning-optimizer

Original：🇺🇸 English
Translated
1 scriptsChecked / no sensitive code detected

Implements the NOWAIT technique for efficient reasoning in R1-style LLMs. Use when optimizing inference of reasoning models (QwQ, DeepSeek-R1, Phi4-Reasoning, Qwen3, Kimi-VL, QvQ), reducing chain-of-thought token usage by 27-51% while preserving accuracy. Triggers on "optimize reasoning", "reduce thinking tokens", "efficient inference", "suppress reflection tokens", or when working with verbose CoT outputs.

31installs
Sourcedavila7/claude-code-templates
Added on

NPX Install

npx skill4agent add davila7/claude-code-templates nowait-reasoning-optimizer

Tags

Translated version includes tags in frontmatter
nowait-techniquellm-inference-optimizationtoken-reductionchain-of-thoughtmultimodal-llms

SKILL.md Content

View Translation Comparison →

NOWAIT Reasoning Optimizer

Implements the NOWAIT technique from the paper "Wait, We Don't Need to 'Wait'! Removing Thinking Tokens Improves Reasoning Efficiency" (Wang et al., 2025).

Overview

NOWAIT is a training-free inference-time intervention that suppresses self-reflection tokens (e.g., "Wait", "Hmm", "Alternatively") during generation, reducing chain-of-thought (CoT) trajectory length by 27-51% without compromising model utility.

When to Use

  • Deploying R1-style reasoning models with limited compute
  • Reducing inference latency for production systems
  • Optimizing token costs for reasoning tasks
  • Working with verbose CoT outputs that need streamlining

Supported Models

Model SeriesTypeToken Reduction
QwQ-32BRL-based16-31%
Phi4-Reasoning-PlusRL-based23-28%
Qwen3-32BRL-based13-16%
Kimi-VL-A3BMultimodal40-60%
QvQ-72B-PreviewMultimodal20-30%
Important: NOWAIT works best with RL-based models. Distilled models (Qwen3-4B/8B/14B) show degraded performance when reflection tokens are suppressed.

Quick Start

1. Basic Implementation

python
from scripts.nowait_processor import NOWAITLogitProcessor

# Initialize processor for your model's tokenizer
processor = NOWAITLogitProcessor(tokenizer)

# Use during generation
outputs = model.generate(
    inputs,
    logits_processor=[processor],
    max_new_tokens=32768
)

2. Keywords Suppressed

See 
references/keywords.md
for the complete list. Core keywords:
wait, alternatively, hmm, but, however, check, 
double-check, maybe, verify, again, oh, ah

How It Works

  1. Initialize Keywords: Identify reflection keywords from empirical analysis
  2. Expand to Token Variants: Map keywords to all token variants in vocabulary (e.g., "wait" → " wait", "Wait", " Wait", ".wait", "WAIT")
  3. Suppress During Inference: Set logits of reflection tokens to large negative values during decoding
Logits (Before)         Logits (After)
Wait     0.8     →     Wait     -inf
First    0.6     →     First    0.6
Hmm      0.5     →     Hmm      -inf
Let      0.4     →     Let      0.4

Key Findings

Why It Works

  • NOWAIT doesn't eliminate self-reflection entirely—it guides models to skip unnecessary "waiting" reasoning
  • Models still perform essential verification at key decision points
  • Results in more linear, straightforward reasoning paths

RL vs Distilled Models

Model TypeNOWAIT EffectRecommendation
RL-based (QwQ, Phi4, Qwen3-32B)Stable accuracy, significant token reduction✅ Recommended
Distilled (Qwen3-4B/8B/14B)Accuracy degradation on hard tasks⚠️ Use with caution
Distilled models rely heavily on CoT structure from training data—removing reflection tokens disrupts their reasoning patterns.

Integration Examples

HuggingFace Transformers

python
from transformers import AutoModelForCausalLM, AutoTokenizer
from scripts.nowait_processor import NOWAITLogitProcessor

model = AutoModelForCausalLM.from_pretrained("Qwen/QwQ-32B")
tokenizer = AutoTokenizer.from_pretrained("Qwen/QwQ-32B")

processor = NOWAITLogitProcessor(tokenizer)

response = model.generate(
    tokenizer(prompt, return_tensors="pt").input_ids,
    logits_processor=[processor],
    max_new_tokens=32768,
    do_sample=True,
    temperature=0.7
)

vLLM

python
from vllm import LLM, SamplingParams
from scripts.nowait_processor import get_nowait_bad_words_ids

llm = LLM(model="Qwen/QwQ-32B")
bad_words_ids = get_nowait_bad_words_ids(llm.get_tokenizer())

sampling_params = SamplingParams(
    max_tokens=32768,
    bad_words_ids=bad_words_ids
)

Expected Results

Task TypeOriginal TokensNOWAIT TokensReduction
Math (AIME)15,00010,50030%
Visual QA (MMMU)2,9001,45050%
Video QA (MMVU)1,7001,25027%

Limitations

  • Less effective on very simple problems where CoT overhead is already minimal
  • Distilled models may suffer accuracy loss on challenging tasks
  • Some domains may require model-specific keyword tuning

References

  • Paper: arXiv:2506.08343v2
  • Complete keyword list: 
    references/keywords.md
  • Implementation: 
    scripts/nowait_processor.py