Beyond Final Answers: CRYSTAL Benchmark for Transparent Multimodal Reasoning Evaluation

• URL: http://arxiv.org/abs/2603.13099v2
• Date: Mon, 16 Mar 2026 03:28:24 GMT
• Title: Beyond Final Answers: CRYSTAL Benchmark for Transparent Multimodal Reasoning Evaluation
• Abstract summary: We introduce CRYSTAL (Clear Reasoning via Yielded Steps, Traceability, and Logic), a diagnostic benchmark with 6,372 instances that evaluates multimodal reasoning.<n>We propose two complementary metrics: Match F1, which scores step-level precision and recall via semantic similarity matching, and Ordered Match F1, which penalizes disordered reasoning chains.<n>CPR-Curriculum achieves a 32% improvement in Match F1 via GRPO where additive reward strategies fail, improving reasoning without manual step annotation.
• Score: 3.23600523782706
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We introduce CRYSTAL (Clear Reasoning via Yielded Steps, Traceability, and Logic), a diagnostic benchmark with 6,372 instances that evaluates multimodal reasoning through verifiable intermediate steps. We propose two complementary metrics: Match F1, which scores step-level precision and recall via semantic similarity matching, and Ordered Match F1, which further penalizes disordered reasoning chains. References are constructed through a Delphi-inspired pipeline in which four independent MLLMs generate trajectories, which are then aggregated via semantic clustering and validated through human quality gates. Evaluation of 20 MLLMs, including commercial frontier systems not used during benchmark construction, reveals systematic failures that are invisible to answer accuracy: universal cherry-picking (precision far exceeds recall), non-monotonic scaling trade-offs, and disordered reasoning in which no competitive model preserves more than 60% of matched steps in the correct order. Beyond evaluation, we propose the Causal Process Reward (CPR), a multiplicative reward that couples answer correctness with step-level alignment, and CPR-Curriculum, which progressively increases reasoning difficulty during training. CPR-Curriculum achieves a 32% improvement in Match F1 via GRPO where additive reward strategies fail, improving reasoning without manual step annotation.

Related papers

• PRO-Step: Step-level Process Reward Optimization for Retrieval-Augmented Generation [2.198430261120653]
  We introduce PRO-STEP: we train a generative PRM that evaluates logical validity and evidential grounding at each step.<n>Experiments on single and multi-hop QA datasets demonstrate that PRO-STEP achieves the best average EM and F1 across five benchmarks.
  arXiv  Detail & Related papers  (2026-08-31T14:34:10Z)
• Claim-Level Reliability Assessment for Efficient Test-Time Reasoning [9.168249719115845]
  We propose claim-level falsification as a principle for test-time scaling and instantiate it through Claim-Level Reliability Assessment (CLR)<n>CLR reallocates test-time compute from additional solution sampling to targeted verification.
  arXiv  Detail & Related papers  (2026-08-12T12:33:05Z)
• Step-wise Rubric Rewards for LLM Reasoning [72.17879367869503]
  Reinforcement Learning with Verifiable Rewards (RLVR) is widely used to improve reasoning in large language models.<n>We find 18.2% of steps in correct responses are wrong yet positively rewarded.<n>Across six mathematical reasoning benchmarks, SRaR improves average accuracy over RaR by 3.57 points.
  arXiv  Detail & Related papers  (2026-05-17T07:08:14Z)
• When Rubrics Fail: Error Enumeration as Reward in Reference-Free RL Post-Training for Virtual Try-On [31.413458927376016]
  We propose Implicit Error Counting to fill a reference-free setting.<n>Instead of checking what a response gets right against a calibrated rubric, IEC enumerates what it gets wrong.<n>We show that nave explicit enumeration is too noisy for stable optimization.
  arXiv  Detail & Related papers  (2026-03-05T20:20:04Z)
• Truncated Step-Level Sampling with Process Rewards for Retrieval-Augmented Reasoning [32.295907409325615]
  Training large language models to reason with search engines via reinforcement learning is hindered by a credit assignment problem.<n>We propose SLATE, a framework built on two complementary ideas.<n> Experiments on seven QA benchmarks confirm that SLATE consistently outperforms both sparse-reward and process-reward baselines.
  arXiv  Detail & Related papers  (2026-02-26T19:05:40Z)
• Overconfident Errors Need Stronger Correction: Asymmetric Confidence Penalties for Reinforcement Learning [17.384089089363382]
  We identify a root cause that existing methods overlook: the uniform penalization of errors.<n>Current approaches treat all incorrect rollouts within a group identically.<n>We propose the Asymmetric Confidence-aware Error Penalty (ACE)
  arXiv  Detail & Related papers  (2026-02-24T22:46:43Z)
• Beyond Correctness: Learning Robust Reasoning via Transfer [51.403609251508904]
  We adopt a simple philosophical view, robust reasoning should remain useful beyond the mind that produced it.<n>We introduce Reinforcement Learning with Transferable Reward, which operationalizes robustness via transfer reward.<n>Our approach improves sampling consistency while improving final answer accuracy, and it reaches comparable performance in substantially fewer training steps.
  arXiv  Detail & Related papers  (2026-02-09T10:41:44Z)
• VERGE: Formal Refinement and Guidance Engine for Verifiable LLM Reasoning [4.3414302048068745]
  We present a neurosymbolic framework that combines Large Language Models with SMT solvers to produce verification-guided answers.<n>We introduce three key innovations: (1) multi-model consensus via formal semantic equivalence checking, (2) semantic routing that directs different claim types to appropriate verification strategies, and (3) precise logical error localization via Minimal Correction Subsets.<n>With the GPT-OSS-120B model, VERGE demonstrates an average performance uplift of 18.7% at convergence across a set of reasoning benchmarks compared to single-pass approaches.
  arXiv  Detail & Related papers  (2026-01-27T20:59:11Z)
• Save the Good Prefix: Precise Error Penalization via Process-Supervised RL to Enhance LLM Reasoning [59.76691952347156]
  Reinforcement learning (RL) has emerged as a powerful framework for improving the reasoning capabilities of large language models (LLMs)<n>Most existing RL approaches rely on sparse outcome rewards, which fail to credit correct intermediate steps in partially successful solutions.<n>We propose Verifiable Prefix Policy Optimization (VPPO), which uses PRMs only to localize the first error during RL.
  arXiv  Detail & Related papers  (2026-01-26T21:38:20Z)
• CARE What Fails: Contrastive Anchored-REflection for Verifiable Multimodal [84.71254539482369]
  Group-relative reinforcement learning with verifiable rewards (RLVR) often wastes the most informative data it already has the failures.<n>We present CARE, a failure-centric post-training framework for multimodal reasoning that turns errors into supervision.<n> CARE improves accuracy and training smoothness while explicitly increasing the share of learning signal that comes from failures.
  arXiv  Detail & Related papers  (2025-12-22T16:34:21Z)
• RefineBench: Evaluating Refinement Capability of Language Models via Checklists [71.02281792867531]
  We evaluate two refinement modes: guided refinement and self-refinement.<n>In guided refinement, both proprietary LMs and large open-weight LMs can leverage targeted feedback to refine responses to near-perfect levels within five turns.<n>These findings suggest that frontier LMs require breakthroughs to self-refine their incorrect responses.
  arXiv  Detail & Related papers  (2025-11-27T07:20:52Z)
• Enhancing the Outcome Reward-based RL Training of MLLMs with Self-Consistency Sampling [90.87033586963828]
  Outcome-reward reinforcement learning (RL) is a common and increasingly significant way to refine the step-by-step reasoning of multimodal large language models (MLLMs)<n>We propose Self-Consistency Sampling (SCS) to correct this issue.<n>Based on Qwen2.5-VL-7B-Instruct, SCS improves accuracy by up to 7.7 percentage points on six multimodal benchmarks with negligible extra computation.
  arXiv  Detail & Related papers  (2025-11-13T18:59:57Z)
• Eigen-1: Adaptive Multi-Agent Refinement with Monitor-Based RAG for Scientific Reasoning [53.45095336430027]
  We develop a unified framework that combines implicit retrieval and structured collaboration.<n>On Humanity's Last Exam (HLE) Bio/Chem Gold, our framework achieves 48.3% accuracy.<n>Results on SuperGPQA and TRQA confirm robustness across domains.
  arXiv  Detail & Related papers  (2025-09-25T14:05:55Z)
• Mind the Gap: A Closer Look at Tokenization for Multiple-Choice Question Answering with LLMs [16.357595595062946]
  There is no consensus on how to tokenize the space following the colon, often overlooked as a trivial choice.<n>Surprisingly, we are able to recommend one specific strategy -- tokenizing the space together with the answer letter.<n>Our findings underscore the importance of careful evaluation design and highlight the need for standardized, transparent evaluation protocols.
  arXiv  Detail & Related papers  (2025-09-18T14:47:58Z)
• CompassVerifier: A Unified and Robust Verifier for LLMs Evaluation and Outcome Reward [50.97588334916863]
  We develop CompassVerifier, an accurate and robust lightweight verifier model for evaluation and outcome reward.<n>It demonstrates multi-domain competency spanning math, knowledge, and diverse reasoning tasks, with the capability to process various answer types.<n>We introduce VerifierBench benchmark comprising model outputs collected from multiple data sources, augmented through manual analysis of metaerror patterns to enhance CompassVerifier.
  arXiv  Detail & Related papers  (2025-08-05T17:55:24Z)
• GRPO-CARE: Consistency-Aware Reinforcement Learning for Multimodal Reasoning [53.894789613838654]
  We introduce SEED-Bench-R1, a benchmark with complex real-world videos requiring balanced perception and reasoning.<n>Using SEED-Bench-R1, we find that standard GRPO, while improving answer accuracy, often reduces logical coherence between reasoning steps and answers, with only a 57.9% consistency rate.<n>We propose GRPO-CARE, a consistency-aware RL framework optimizing both answer correctness and reasoning coherence without explicit supervision.
  arXiv  Detail & Related papers  (2025-06-19T08:49:13Z)
• Step-KTO: Optimizing Mathematical Reasoning through Stepwise Binary Feedback [94.25162866972077]
  Step-KTO is a training framework that combines process-level and outcome-level binary feedback.<n>Our experiments show that Step-KTO significantly improves both final answer accuracy and the quality of intermediate reasoning steps.
  arXiv  Detail & Related papers  (2025-01-18T15:38:03Z)

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.