AdaRubric: Task-Adaptive Rubrics for LLM Agent Evaluation

• URL: http://arxiv.org/abs/2603.21362v1
• Date: Sun, 22 Mar 2026 18:47:34 GMT
• Title: AdaRubric: Task-Adaptive Rubrics for LLM Agent Evaluation
• Abstract summary: LLM-as-Judge evaluation fails agent tasks because a fixed rubric cannot capture what matters for this task.<n>We present ADARUBRIC, which generates task-specific evaluation rubrics on the fly from task descriptions.<n>On WebArena and ToolBench, ADARUBRIC achieves human correlation (+0.16 over the best static baseline) with deployment-grade reliability.
• Score: 12.42698406379544
• License: http://creativecommons.org/publicdomain/zero/1.0/
• Abstract: LLM-as-Judge evaluation fails agent tasks because a fixed rubric cannot capture what matters for this task: code debugging demands Correctness and Error Handling; web navigation demands Goal Alignment and Action Efficiency. We present ADARUBRIC, which closes this gap by generating task-specific evaluation rubrics on the fly from task descriptions, scoring trajectories step-by-step with confidence-weighted per-dimension feedback, and filtering preference pairs with the novel DimensionAwareFilter - a provably necessary condition for preventing high-scoring dimensions from masking dimension-level failures. On WebArena and ToolBench, ADARUBRIC achieves Pearson r=0.79 human correlation (+0.16 over the best static baseline) with deployment-grade reliability (Krippendorff's $α$=0.83). DPO agents trained on ADARUBRIC preference pairs gain +6.8 to +8.5 pp task success over Prometheus across three benchmarks; gains transfer to SWE-bench code repair (+4.9 pp) and accelerate PPO convergence by +6.6 pp at 5K steps - both without any rubric engineering. Code: https://github.com/alphadl/AdaRubrics.

Related papers

• ReDraft, Don't Just Distill: Reference-Driven Revision for Continual VLLM Post-Training [66.72708893922605]
  Post-training of large multimodal models should add new capabilities while preserving those from pre-training.<n>SFT gives explicit target supervision that learns a task from near-zero accuracy, but its off-policy targets move the model far enough to cause forgetting.<n>We introduce ReDraft, which obtains both from the model's own failures.
  arXiv  Detail & Related papers  (2026-09-15T04:59:03Z)
• DPIAgent: Divide, Protocol, Isolate for Agentic Reproduction Test Generation [31.816679197888366]
  DPIAgent is a structured agentic framework built on three principles.<n>It Divides the task into single-objective phases of defect exploration and test generation.<n>It enforces a handoff Protocol that records the diagnosis and test plan, preventing context loss.
  arXiv  Detail & Related papers  (2026-08-24T14:51:15Z)
• Navigation-Informed Embeddings: Dense-Retriever Adaptation from Agent Search Traces [0.0]
  We introduce Navigation-Informed Embeddings (NIE), a family of trace-derived objectives.<n>NIE-Stop turns the stopping document into a soft positive; NIE-Path additionally uses preceding path documents as hard comparisons.<n>NIE-Path raises long-path performance to 55.4, compared with 46.7 for the unadapted encoder.
  arXiv  Detail & Related papers  (2026-08-16T23:07:44Z)
• AgentDebugX: An Open-Source Toolkit for Failure Observability, Attribution, and Recovery in LLM Agents [83.37969790806069]
  LLM agent failures are difficult to debug because the step where an error surfaces is often not the one that caused it.<n>We present Agent DebuggerX, an open-source framework that organizes debug as a closed loop of Detect, Attribute, Recover, and Rerun.<n>At its core, Deep Debugger performs multi-turn root-cause diagnosis through global trajectory understanding.
  arXiv  Detail & Related papers  (2026-07-21T06:21:13Z)
• Amplitude-Only FFN Intervention for Tool-Structured LLM Inference Method: Gated Evaluation Protocol, and Cross-Model Empirical Results [10.526208975530054]
  We study inference-time feed-forward network (FFN) intervention for improving structured outputs without retraining model weights.<n>Amplitude Gating (AG) is a non-destructive alternative that preserves pretrained FFN weight directions and modulates only activation magnitudes during generation.<n>These results identify tool-structured inference as the most credible first target for safe FFN-level inference optimization.
  arXiv  Detail & Related papers  (2026-07-13T07:28:55Z)
• Efficient Visual Pointing for Embodied AI:Agent-Driven Data Synthesis, Cross-Block Attention, and Iterative Correction [55.11480729304395]
  We describe our PointArena 2026 solution, which achieves 77.2% overall accuracy and ranks second on the benchmark.<n>The ap proach targets three failure modes. First, agent-driven syn thesis builds large semantic and anchor-relative candidate pools.<n>Second, a determinis tic steerable-data pipeline creates a verified 10,000-sample main set, plus reserve samples, using masks, templates, and path verification.
  arXiv  Detail & Related papers  (2026-06-29T06:39:03Z)
• Trace2Policy: From Expert Behavior Traces to Self-Evolving Decision Agents [5.689042186242701]
  Decision rules that enterprise experts apply tacitly can be systematically recovered and improved through iterative error analysis.<n>We present textbfTrace2Policy, whose core mechanism is textbfEISR.<n>Each round executes the rules on a validation set, clusters errors by root cause into MISSING, WRONG, or CONFLICT types, applies targeted patches, and commits only those that pass a regression gate.
  arXiv  Detail & Related papers  (2026-06-09T06:05:29Z)
• Guardrails as Infrastructure: Policy-First Control for Tool-Orchestrated Workflows [0.0]
  Policy-First Tooling mediates tool invocation through explicit constraints, risk-aware gating, recovery controls, and auditable explanations.<n>In 225 controlled runs across five policy packs and three fault profiles, stricter packs improve violation prevention from 0.000 in P0 to 0.681 in P4, while task success drops from 0.356 to 0.067.
  arXiv  Detail & Related papers  (2026-03-18T01:19:33Z)
• Surgical Post-Training: Cutting Errors, Keeping Knowledge [14.639241720055658]
  Surgical Post-Training is a paradigm designed to optimize reasoning efficiently while preserving learned prior knowledge.<n>With only 4k rectified math data pairs, SPoT improves Qwen3-8B's accuracy by 6.2% on average across in-domain and OOD tasks.
  arXiv  Detail & Related papers  (2026-03-02T10:12:56Z)
• CLEANER: Self-Purified Trajectories Boost Agentic Reinforcement Learning [4.765206163164323]
  CLEANER exploits intrinsic self-correction capabilities to eliminate error-contaminated context during data collection.<n>Similarity-Aware Adaptive Rollback mechanism autonomously constructs clean, purified trajectories.<n>Results show average accuracy gains of 6%, 3%, and 5% over baselines.
  arXiv  Detail & Related papers  (2026-01-21T16:14:30Z)
• SABER: Small Actions, Big Errors -- Safeguarding Mutating Steps in LLM Agents [52.20768003832476]
  We analyze execution traces on $$-Bench (Airline/Retail) and SWE-Bench Verified.<n>We formalize emphdecisive deviations, earliest action, level divergences that flip success to failure.<n>We introduce cm, a model-agnostic, gradient-free, test-time safeguard.
  arXiv  Detail & Related papers  (2025-11-26T01:28:22Z)
• Structured Uncertainty guided Clarification for LLM Agents [126.26213027785813]
  LLM agents extend large language models with tool-calling capabilities, but ambiguous user instructions often lead to incorrect invocations and task failures.<n>We introduce a principled formulation of structured uncertainty over tool-call parameters, modeling joint tool-argument clarification as a POMDP with Expected Value of Perfect Information (EVPI) objective for optimal question selection and aspect-based cost modeling to prevent redundancy.<n>Our SAGE-Agent leverages this structured uncertainty to achieve superior efficiency: increasing coverage on ambiguous tasks by 7-39% while reducing clarification questions by 1.5-2.7$times$ compared to strong prompting and uncertainty-based baselines.
  arXiv  Detail & Related papers  (2025-11-11T21:50:44Z)
• AgentChangeBench: A Multi-Dimensional Evaluation Framework for Goal-Shift Robustness in Conversational AI [5.165179548592513]
  AgentChangeBench is a benchmark designed to measure how tool augmented language model agents adapt to mid dialogue goal shifts.<n>Our framework formalizes evaluation through four complementary metrics: Task Success Rate (TSR) for effectiveness, Tool Use Efficiency (TUE) for reliability, Tool Call Redundancy Rate (TCRR) for wasted effort, and GoalShift Recovery Time (GSRT) for adaptation.
  arXiv  Detail & Related papers  (2025-10-20T23:48:07Z)
• Weights-Rotated Preference Optimization for Large Language Models [30.25242193651982]
  We propose a novel Weights-Rotated Preference Optimization (RoPO) algorithm, which implicitly constrains the output layer logits with the KL divergence inherited from DPO.<n>Our RoPO achieves up to a 3.27-point improvement on AlpacaEval 2, and surpasses the best baseline by 6.2 to 7.5 points on MT-Bench with merely 0.015% of the trainable parameters.
  arXiv  Detail & Related papers  (2025-08-25T03:57:17Z)
• SOPBench: Evaluating Language Agents at Following Standard Operating Procedures and Constraints [59.645885492637845]
  SOPBench is an evaluation pipeline that transforms each service-specific SOP code program into a directed graph of executable functions.<n>Our approach transforms each service-specific SOP code program into a directed graph of executable functions and requires agents to call these functions based on natural language SOP descriptions.<n>We evaluate 18 leading models, and results show the task is challenging even for top-tier models.
  arXiv  Detail & Related papers  (2025-03-11T17:53:02Z)
• Teaching Your Models to Understand Code via Focal Preference Alignment [70.71693365502212]
  In existing approaches, a set of n candidate solutions is evaluated based on test case success rates.<n>Because this approach aligns entire failing code blocks rather than pinpointing specific errors, it lacks the granularity necessary to capture meaningful error-correction relationships.<n>We propose Target-DPO, a new preference alignment framework that mimics human iterative debug to refine Code LLMs.
  arXiv  Detail & Related papers  (2025-03-04T16:56:34Z)
• Step-DPO: Step-wise Preference Optimization for Long-chain Reasoning of LLMs [54.05511925104712]
  We propose a simple, effective, and data-efficient method called Step-DPO. Step-DPO treats individual reasoning steps as units for preference optimization rather than evaluating answers holistically. Our findings demonstrate that as few as 10K preference data pairs and fewer than 500 Step-DPO training steps can yield a nearly 3% gain in accuracy on MATH for models with over 70B parameters.
  arXiv  Detail & Related papers  (2024-06-26T17:43:06Z)

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.