Asking LLMs to Verify First is Almost Free Lunch
Department of Electronic Engineering, Tsinghua University
Abstract
To enhance the reasoning capabilities of Large Language Models (LLMs) without high costs of training, nor extensive test-time sampling, we introduce Verification-First (VF), a strategy that prompts models to verify a provided candidate answer—even a trivial or random one—before generating a solution. This approach triggers a "reverse reasoning" process that is cognitively easier and complementary to standard forward Chain-of-Thought (CoT), effectively invoking the model's critical thinking to reduce logical errors.
Theoretical Foundations
Verifying an answer is easier than generating a correct answer, providing complementary information to standard CoT
Asking one to criticize an answer from others can invoke critical thinking by overcoming egocentrism
Methodology
A. Let's think step by step..."✅
A possible answer of Q is A'. First verify if A' is correct, then think step by step to find the answer."
Iter-VF Process: Iterative application of VF using the model's previous answer, creating a Markovian refinement loop that avoids context length and error propagation issues.
Experimental Results
Performance comparison across various benchmarks
| Method | Output Tokens (Relative to CoT) | Performance Gain |
|---|---|---|
| Standard CoT | 100% | Baseline |
| Verification-First | 120-150% | +15% to +25% |
| Self-Consistency (N=5) | 500% | +10% to +20% |
Key Findings
- VF with random answers consistently outperforms standard CoT with minimal computational overhead
- Iter-VF outperforms existing TTS strategies under limited computational budgets
- VF is effective even with thought-hidden commercial LLM services
- Verification process is the key driver of improvement, not the quality of the candidate answer