Reasoning Is Not Free: Robust Adaptive Cost-Efficient Routing for LLM-as-a-Judge

• URL: http://arxiv.org/abs/2605.10805v1
• Date: Mon, 11 May 2026 16:30:20 GMT
• Title: Reasoning Is Not Free: Robust Adaptive Cost-Efficient Routing for LLM-as-a-Judge
• Abstract summary: Reasoning-capable large language models (LLMs) have recently been adopted as automated judges.<n>We show that explicit reasoning substantially improves judgment accuracy on tasks requiring structured verification.<n>We propose a Robust Adaptive Cost-Efficient Routing (RACER) which selects between reasoning and non-reasoning judges.
• Score: 4.511996087821266
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Reasoning-capable large language models (LLMs) have recently been adopted as automated judges, but their benefits and costs in LLM-as-a-Judge settings remain unclear. Through controlled comparisons between reasoning and non-reasoning judges, we show that explicit reasoning substantially improves judgment accuracy on tasks requiring structured verification (e.g., math and coding), while offering limited or even negative gains on simpler evaluations and incurring significantly higher computational cost. These findings motivate that reasoning should be used selectively rather than universally, with awareness of possible distribution shift. We propose a Robust Adaptive Cost-Efficient Routing (RACER), which dynamically selects between reasoning and non-reasoning judges under a fixed budget by formulating routing as a constrained distributionally robust optimization problem. RACER explicitly accounts for distribution shift via a KL-divergence uncertainty set, admits an efficient primal--dual algorithm, and enjoys theoretical guarantees including uniqueness of the optimal policy and linear convergence. Extensive experiments show that RACER achieves superior accuracy--cost trade-offs under distribution shift.

Related papers

• Joint Optimization of Human Headcount and Stochastic AI Resource Capacity [0.0]
  Leadership of organizations must split a fixed budget between human headcount and generative-AI token capacity.<n>Standard deterministic planning ignores both the heavy-tailed volatility of token consumption and the cognitive cost of auditing AI-generated output.<n>We show that rising individual token volatility raises optimal per-capita token intensity while contracting headcount.
  arXiv  Detail & Related papers  (2026-08-01T22:34:02Z)
• TRACE: A Unified Rollout Budget Allocation Framework for Efficient Agentic Reinforcement Learning [52.67756371231985]
  Reinforcement learning with verifiable rewards (RLVR) is a promising approach for enhancing reasoning and agentic behavior in large language models.<n>We introduce Tree Rollout Allocation for Contrastive Exploration (TRACE), a unified rollout allocation framework.<n>Technically, TRACE allocates rollout budget to both prompt roots and intermediate prefixes that are most likely to yield mixed terminal rewards.
  arXiv  Detail & Related papers  (2026-06-09T17:16:03Z)
• The Shadow Price of Reasoning: Economic Perspective on Optimal Budget Allocation for LLMs [12.942771885641719]
  Inference-time scaling has emerged as a critical avenue for enhancing Large Language Models' performance.<n>In this work, we formulate inference budget allocation as a global constrained optimization problem governed by economic principles.
  arXiv  Detail & Related papers  (2026-06-02T03:26:55Z)
• Instance-Optimal Estimation with Multiple LLM Judges on a Budget [84.31744861038106]
  We formalize this question as *budgeted heteroskedastic multi-judge estimation*.<n>Given $K$ prompt-response pairs, $J$ judges with known costs, and unknown query-judge variances, the goal is to estimate a bounded score vector while minimizing an $ell_p$-error.<n>We prove that EST-IVWE matches the oracle IVWE rate up to lower-order terms in the budget.
  arXiv  Detail & Related papers  (2026-05-22T08:26:08Z)
• ODAR: Principled Adaptive Routing for LLM Reasoning via Active Inference [60.958331943869126]
  ODAR-Expert is an adaptive routing framework that optimize the accuracy-efficiency trade-off via principled resource allocation.<n>We show strong and consistent gains, including 98.2% accuracy on MATH and 54.8% on Humanity's Last Exam.
  arXiv  Detail & Related papers  (2026-02-27T05:22:01Z)
• WS-GRPO: Weakly-Supervised Group-Relative Policy Optimization for Rollout-Efficient Reasoning [67.45237332694025]
  Group Relative Policy Optimization is effective for training language models on complex reasoning.<n>We propose Weakly Supervised GRPO, which improves rollout efficiency by converting terminal rewards into correctness-aware guidance.
  arXiv  Detail & Related papers  (2026-02-19T02:43:35Z)
• Conformal Thinking: Risk Control for Reasoning on a Compute Budget [60.65072883773352]
  Reasoning Large Language Models (LLMs) enable test-time scaling, with dataset-level accuracy improving as the token budget increases.<n>We re-frame the budget setting problem as risk control, limiting the error rate while minimizing compute.<n>Our framework introduces an upper threshold that stops reasoning when the model is confident and a novel lower threshold that preemptively stops unsolvable instances.
  arXiv  Detail & Related papers  (2026-02-03T18:17:22Z)
• AdaGReS:Adaptive Greedy Context Selection via Redundancy-Aware Scoring for Token-Budgeted RAG [4.69377227249912]
  We present AdaGReS, a redundancy-aware context selection framework for token-budgeted RAG.<n>AdaGReS performs greedy selection under a token-budget constraint using marginal gains derived from the objective.<n> Experiments on open-domain question answering (Natural Questions) and a high-redundancy biomedical (drug) corpus demonstrate consistent improvements in redundancy control and context quality.
  arXiv  Detail & Related papers  (2025-12-31T18:48:07Z)
• Additive Distributionally Robust Ranking and Selection [0.8283940114367679]
  We propose a simple additive allocation (AA) procedure that aims to exclusively sample the $k + m - 1$ previously hypothesized critical scenarios.<n>We prove that AA is consistent and, surprisingly, achieves additivity in the strongest sense.<n>Results offer new and counterintuitive insights into the additive structure of DRR&S.
  arXiv  Detail & Related papers  (2025-09-07T17:36:29Z)
• Supervised Optimism Correction: Be Confident When LLMs Are Sure [91.7459076316849]
  We establish a novel theoretical connection between supervised fine-tuning and offline reinforcement learning.<n>We show that the widely used beam search method suffers from unacceptable over-optimism.<n>We propose Supervised Optimism Correction, which introduces a simple yet effective auxiliary loss for token-level $Q$-value estimations.
  arXiv  Detail & Related papers  (2025-04-10T07:50:03Z)
• Reasoning Aware Self-Consistency: Leveraging Reasoning Paths for Efficient LLM Sampling [9.44858963874474]
  Self-Consistency mitigates hallucinations in Large Language Models (LLMs) by sampling multiple reasoning paths.<n>We introduce Reasoning-Aware Self-Consistency (RASC), a novel framework that enhances sampling efficiency and reasoning faithfulness.
  arXiv  Detail & Related papers  (2024-08-30T05:14:59Z)
• Provably Mitigating Overoptimization in RLHF: Your SFT Loss is Implicitly an Adversarial Regularizer [52.09480867526656]
  We identify the source of misalignment as a form of distributional shift and uncertainty in learning human preferences.<n>To mitigate overoptimization, we first propose a theoretical algorithm that chooses the best policy for an adversarially chosen reward model.<n>Using the equivalence between reward models and the corresponding optimal policy, the algorithm features a simple objective that combines a preference optimization loss and a supervised learning loss.
  arXiv  Detail & Related papers  (2024-05-26T05:38:50Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.