DiSCTT: Consensus-Guided Self-Curriculum for Efficient Test-Time Adaptation in Reasoning

2026-03-05

AI Summary

This work addresses the inefficiency and training instability of existing test-time adaptation methods in heterogeneous reasoning tasks, which stem from their use of a uniform optimization objective. To overcome these limitations, we propose a self-curriculum framework guided by instance difficulty and consensus. Sample difficulty is dynamically assessed through the consistency of reasoning trajectories: high-consensus instances are refined via pseudo-label supervised fine-tuning, while low-consensus instances are optimized using reinforcement learning augmented with consensus-based regularization. Our approach is the first to integrate instance-level epistemic uncertainty and consensus mechanisms into test-time adaptation, enabling difficulty-aware dynamic strategy allocation. Experiments across multiple mathematical and general reasoning benchmarks demonstrate significant improvements over state-of-the-art methods, achieving higher accuracy, lower variance, and substantially reduced computational cost and training time.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Test-time adaptation offers a promising avenue for improving reasoning performance in large language models without additional supervision, but existing approaches often apply a uniform optimization objective across all inputs, leading to inefficient or unstable adaptation on heterogeneous reasoning problems. We propose DiSCTT, a difficulty-aware, consensus-guided self-curriculum framework that dynamically allocates test-time optimization strategies based on instance-level epistemic uncertainty estimated from agreement among sampled reasoning trajectories. Inputs with high consensus are consolidated via supervised fine-tuning using majority-agreed solutions as pseudo-labels, while low-consensus inputs are optimized via reinforcement learning with a consensus-regularized objective that encourages diversity under relevance constraints. Across a broad suite of mathematical and general reasoning benchmarks, DiSCTT consistently outperforms strong test-time adaptation baselines, achieving higher accuracy with reduced variance and substantially lower computation and wall-clock training times. These results demonstrate that explicitly accounting for instance difficulty and uncertainty enables more stable, efficient, and effective test-time adaptation for reasoning models.

Problem

• test-time adaptation
• reasoning
• heterogeneous reasoning problems
• optimization efficiency
• adaptation stability

Innovation

• test-time adaptation
• self-curriculum
• consensus-guided
• epistemic uncertainty
• reasoning