When Self-Consistency Hurts: The Hidden Cost of Majority Voting in Small Language Models
cs.AI
Submitted: 2026-11-08
Updated: 2026-11-08
Code: https://github.com/heywanrong/SiftSC
Terminology
Sources
- Let's Sample Step by Step: Adaptive-Consistency for Efficient Reasoning and Coding with LLMs
- CGES: Confidence-Guided Early Stopping for Efficient and Accurate Self-Consistency
- Mobile-MMLU: A Mobile Intelligence Language Understanding Benchmark
- Task-Specific Efficiency Analysis: When Small Language Models Outperform Large Language Models
- Are More LLM Calls All You Need? Towards Scaling Laws of Compound Inference Systems
- Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models
- Learning to Route LLMs with Confidence Tokens
- Training Verifiers to Solve Math Word Problems
- Certified Self-Consistency: Statistical Guarantees and Test-Time Training for Reliable Reasoning in LLMs
- Gemma 3 Technical Report
- Seer Self-Consistency: Advance Budget Estimation for Adaptive Test-Time Scaling
- T1: Tool-integrated Verification for Test-time Compute Scaling in Small Language Models
- Reliability-Aware Adaptive Self-Consistency for Efficient Sampling in LLM Reasoning
- Quantization Meets Reasoning: Exploring LLM Low-Bit Quantization Degradation for Mathematical Reasoning
- Let's Verify Step by Step
- PETS: A Principled Framework Towards Optimal Trajectory Allocation for Efficient Test-Time Self-Consistency
- RouteLLM: Learning to Route LLMs with Preference Data
- OpenAI o1 System Card
- Qwen2.5 Technical Report
- Token Level Routing Inference System for Edge Devices
Related papers
- Multi-Objective Structured Pruning of LLMs for Latency and Model Size Optimization
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Embeddings for Preferences, Not Semantics
- Can We Trust Item Response Theory for AI Evaluation?
- Diagnosing Task Insensitivity in Language Agents