Judge Circuits

• URL: http://arxiv.org/abs/2605.16023v1
• Date: Fri, 15 May 2026 14:57:21 GMT
• Title: Judge Circuits
• Abstract summary: We investigate the internal mechanism in Gemma-3, Qwen2.5, and Llama-3.<n>We find that judgments across structured understanding and open-ended preference tasks share a sparse, generalized Latent Evaluator sub-graph.<n>We provide a mechanistic account of format-induced inconsistency on the open-weight models we study.
• Score: 24.717818484378583
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: LLM-as-a-judge has become the dominant paradigm for grading model outputs at scale, yet the same model assigns systematically different scores when its output format changes (e.g., a 1-5 rating vs. a True/False label). Existing diagnoses of these format-induced inconsistencies stop at the input-output level. Using Position-aware Edge Attribution Patching (PEAP), we causally investigate the internal mechanism in Gemma-3, Qwen2.5, and Llama-3. We find that judgments across structured understanding and open-ended preference tasks share a sparse, generalized Latent Evaluator sub-graph in the mid-to-late multi-layer perceptrons (MLPs); zero-ablating it collapses judgment while preserving world knowledge in architecturally modular models. By structurally decoupling abstract judging from output formatting, we provide a mechanistic account of format-induced inconsistency on the open-weight models we study: a continuous judgment signal computed in the shared trunk is mapped through fragile, format-specific terminal branches, enabling format-independent preference to be isolated downstream of the requested output format. Our findings imply that benchmark-level reliability comparisons across formats are partially measuring formatter geometry rather than evaluation quality.

Related papers

• The RAT: A Unified Bayesian Model for RAG Evaluation [8.650407684623714]
  We introduce a Bayesian evaluation framework that jointly models retrieval success, abstention behavior, and answer correctness.<n>We apply the framework to 27 RAG configurations across three datasets, three retrievers, and three generators.<n>We extend the model to incorporate LLM-as-a-judge annotations as calibrated noisy observations.
  arXiv  Detail & Related papers  (2026-08-25T15:54:30Z)
• A Unified Backbone--Expert Framework with Relation-Token and Residual--Classifier Interfaces for Automatic Modulation Recognition [1.547549252134621]
  We propose a unified backbone-expert framework with a common convolutional state-space backbone and two specialized interfaces.<n>Our framework achieves overall average accuracies of 67.28 pm 0.14% on RML2016.10b and 87.19 pm 0.77% on HisarMod 2019.
  arXiv  Detail & Related papers  (2026-08-15T10:26:42Z)
• DriftScope: Measuring The Hidden Effects of Diffusion Model Adaptation [50.584834483403675]
  Adapting pre-trained text-to-image diffusion models, whether to learn new visual concepts or erase unwanted ones, is routinely evaluated on its intended effects alone.<n>We demonstrate that adaptation systematically damages semantically unrelated concepts in ways that aggregate metrics structurally cannot surface.<n>We introduce DriftScope, a prompt-level diagnostic tool that takes any two model checkpoints and returns a ranked list of tokens whose visual concepts have shifted most between them.
  arXiv  Detail & Related papers  (2026-06-30T20:57:51Z)
• DualEval: Joint Model-Item Calibration for Unified LLM Evaluation [67.99119926012686]
  DualEval is a latent model-item calibration framework that represents models and evaluation items in a shared space.<n>We apply DualEval across four domains: coding, math, miscellaneous domain-knowledge tasks, and generic everyday user queries.
  arXiv  Detail & Related papers  (2026-06-24T22:40:46Z)
• Gatekeepers and Hallucinations: A Layered Evaluation Framework for LLM-Driven Quantum Circuit Generation [0.0]
  We present a layered evaluation framework for materials-informed Variational Quantum Eigensolver (VQE) circuit generation.<n>We surface a taxonomy of five distinct LLM failure modes (geometry hallucination, nonexistent API usage, integration failures, constraint violations, and plausible-but-unverifiable output)<n>A forensic audit of the evaluation platform's own source code further establishes that two apparent model failures originated in the harness through silent fallback-template substitution.
  arXiv  Detail & Related papers  (2026-06-16T19:17:54Z)
• Deployment-Relevant Alignment Cannot Be Inferred from Model-Level Evaluation Alone [11.663456969895462]
  Alignment evaluation in machine learning has largely become evaluation of models.<n>This paper argues that deployment-relevant alignment cannot be inferred from model-level evaluation alone.
  arXiv  Detail & Related papers  (2026-05-06T03:28:30Z)
• Benchmarking Layout-Guided Diffusion Models through Unified Semantic-Spatial Evaluation in Closed and Open Settings [5.994477544297929]
  We introduce a closed-set benchmark (C-Bench) to isolate key generative capabilities while providing varying levels of complexity in both prompt structure and layout.<n>To complement this controlled setting, we propose an open-set benchmark (O-Bench) that evaluates models using real-world prompts and layouts.<n>We conduct a large-scale evaluation of six state-of-the-art layout-guided diffusion models, totaling 319,086 generated and evaluated images.
  arXiv  Detail & Related papers  (2026-04-28T08:25:32Z)
• The Last Fingerprint: How Markdown Training Shapes LLM Prose [0.0]
  We propose that the em dash is markdown leaking into prose.<n>We present a five-step genealogy connecting training data composition, structural internalization, the dual-register status of the em dash, and post-training amplification.
  arXiv  Detail & Related papers  (2026-03-27T21:42:06Z)
• FMBench: Adaptive Large Language Model Output Formatting [49.52930069696333]
  We present FMBench, a benchmark for adaptive Markdown output formatting.<n>Experiments on two model families show that SFT consistently improves semantic alignment.<n>Results also reveal an inherent trade-off between semantic and structural objectives.
  arXiv  Detail & Related papers  (2026-02-06T04:42:06Z)
• Dependence-Aware Label Aggregation for LLM-as-a-Judge via Ising Models [55.94503936470247]
  Large-scale AI evaluation increasingly relies on aggregating binary judgments from $K$ annotators, including judges.<n>Most classical methods assume annotators are conditionally independent given the true label $Yin0,1$, an assumption often violated by LLM judges.<n>We study label aggregation through a hierarchy of dependence-aware models based on Ising graphical models and latent factors.
  arXiv  Detail & Related papers  (2026-01-29T21:26:50Z)
• SoftPQ: Robust Instance Segmentation Evaluation via Soft Matching and Tunable Thresholds [0.0]
  We propose SoftPQ, a flexible and interpretable instance segmentation metric.<n>We show that SoftPQ captures meaningful differences in segmentation quality that existing metrics overlook.
  arXiv  Detail & Related papers  (2025-05-17T22:08:33Z)
• LLMs Are Biased Towards Output Formats! Systematically Evaluating and Mitigating Output Format Bias of LLMs [69.40865293066885]
  We present the first systematic evaluation examining format bias in performance of large language models (LLMs)<n>We present our empirical format bias evaluation spanning four commonly used categories -- multiple-choice question-answer, wrapping, list, and mapping.
  arXiv  Detail & Related papers  (2024-08-16T10:45:45Z)
• Sparse Conditional Hidden Markov Model for Weakly Supervised Named Entity Recognition [68.68300358332156]
  We propose the sparse conditional hidden Markov model (Sparse-CHMM) to evaluate noisy labeling functions. Sparse-CHMM is optimized through unsupervised learning with a three-stage training pipeline. It achieves a 3.01 average F1 score improvement on five comprehensive datasets.
  arXiv  Detail & Related papers  (2022-05-27T20:47:30Z)
• How Faithful is your Synthetic Data? Sample-level Metrics for Evaluating and Auditing Generative Models [95.8037674226622]
  We introduce a 3-dimensional evaluation metric that characterizes the fidelity, diversity and generalization performance of any generative model in a domain-agnostic fashion. Our metric unifies statistical divergence measures with precision-recall analysis, enabling sample- and distribution-level diagnoses of model fidelity and diversity.
  arXiv  Detail & Related papers  (2021-02-17T18:25:30Z)
• Autoencoding Variational Autoencoder [56.05008520271406]
  We study the implications of this behaviour on the learned representations and also the consequences of fixing it by introducing a notion of self consistency. We show that encoders trained with our self-consistency approach lead to representations that are robust (insensitive) to perturbations in the input introduced by adversarial attacks.
  arXiv  Detail & Related papers  (2020-12-07T14:16:14Z)

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.