FlavourBench: Executable Culinary Reward Maps for Language Model Evaluation and Post-Training

• URL: http://arxiv.org/abs/2608.20574v2
• Date: Wed, 26 Aug 2026 20:41:39 GMT
• Title: FlavourBench: Executable Culinary Reward Maps for Language Model Evaluation and Post-Training
• Abstract summary: Open-ended language-model evaluation often substitutes another model or a small preference panel for a missing answer key.<n>We introduce FlavourBench, which compiles dense answer maps from a versioned culinary environment.<n>We evaluate 27 frontier endpoints on the same 534 substitution, pairing, and constraint tasks, yielding 14,418 complete model-task observations.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Open-ended language-model evaluation often substitutes another model or a small preference panel for a missing answer key. We introduce FlavourBench, which instead compiles dense answer maps from a versioned culinary environment. Each task asks for a three-ingredient portfolio from eight candidates; before inference, Epicure scores all 56 portfolios. We evaluate 27 frontier endpoints on the same 534 substitution, pairing, and constraint tasks, yielding 14,418 complete model-task observations. Anchor-cluster bootstraps and multiplicity-controlled paired tests resolve 101 of 351 model contrasts. Grok 4.6 has the largest point estimate at 65.1, but the corrected evidence does not identify a unique best endpoint. The ranking replicates across independently compiled panels and remains similar under alternative metrics, task filters, family weights, and three public Epicure checkpoints. We then run a preregistered, three-seed post-training study. LoRA SFT of a pinned Qwen3-0.6B checkpoint on 270 Epicure-optimal answers improves its score on 84 anchor-disjoint maps by 13.30 points over a format- and label-matched control (95% CI 6.52 to 20.29, p = 0.000170), and the effect replicates on all 534 public maps. The replication gain is 11.73 points (95% CI 8.98 to 14.54). On the primary split, both trained arms parse every response while the format control does not improve on the base model. The release contains prompts, exhaustive reward maps, raw responses, training and evaluation manifests, statistical plans, code, and an offline verifier.

Related papers

• Coding Agents Have Converged: Why the SWE-bench Leaderboard Can No Longer Order Its Top Entries, and What to Measure Instead [9.25138311822007]
  Small differences on coding-agent leaderboards are often read as an ordering of systems.<n>We audit whether the published verdicts support this reading, using 254 SWE-bench submissions across four splits without running models.<n>We release the partition and a five-step audit protocol that profiles shared outcomes, tests paired differences, reports grouping sensitivity, and estimates the instance budget needed for resolution.
  arXiv  Detail & Related papers  (2026-09-15T16:29:12Z)
• MetroLLM-Bench: Evaluating Language Models as Transit Kiosk Runtimes [0.0]
  We introduce MetroLLM-Bench, a 955-case benchmark for testing language models as the policy layer of a transit kiosk.<n>It covers six real metro systems, ranging from 37 to 414 stations, and eleven categories that include routing, fare calculation, disruptions, accessibility, and adversarial input.<n>14 deterministic scoring components form Tier 1; eight semantic-quality components form Tier 2, six of which use a language-model judge.
  arXiv  Detail & Related papers  (2026-09-09T10:46:56Z)
• TRACE-Bench: Decomposing and Diagnosing Multi-Reference Image Generation [74.69068352546073]
  TRACE-Bench is an evaluation tool for multi-reference images spanning diverse artistic styles and real-world subjects.<n>It is built from 631 formula templates and around 4,000 reference images spanning diverse artistic styles and real-world subjects.<n>Its formula structure drives an operator-aligned evaluation protocol for per-capability scoring and a diagnostic tree analysis.
  arXiv  Detail & Related papers  (2026-08-17T16:15:50Z)
• 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)
• When Does Combining Language Models Help? A Co-Failure Ceiling on Routing, Voting, and Mixture-of-Agents Across 67 Frontier Models [0.0]
  Multi-model LLM systems are used to beat single-model accuracy.<n>We show that their gain is capped by a quantity the field rarely reports.
  arXiv  Detail & Related papers  (2026-06-25T17:06:06Z)
• You Don't Need to Run Every Eval [14.183605234187624]
  A model's scores across all 133 benchmarks are largely determined by just two numbers.<n>We release the score matrix, the BenchPress code, and an interactive tool that predicts any model's score on any benchmark.
  arXiv  Detail & Related papers  (2026-06-22T23:54:00Z)
• CoEval: Ranking Language Models for Custom Tasks Without Labeled Data or Trustworthy Benchmarks [47.027290803102666]
  We present CoEval, an open framework that supplies a trustworthy, task-specific signal through ensemble self-evaluation.<n>A pool of models rotates through all three roles, teacher, student, and judge, to generate a fresh, contamination-free benchmark.
  arXiv  Detail & Related papers  (2026-06-02T13:41:43Z)
• Right Makes Might: Aligning Verified Hidden States Empowers RL Reasoning [55.264863369127774]
  Current methods reduce each correct rollout to a single reward bit, ignoring the geometric structure shared among their hidden states.<n>We propose Hidden-Align, an auxiliary loss function that aligns the last-layer hidden states of correct rollouts at the anchor token during RL training, with zero overhead in both training and inference.<n>On eight mathematical reasoning benchmarks, Hidden-Align improves average pass@1 over the DAPO baseline by 3.8, 6.2, and 5.4 percentage points on Qwen3-1.7B, 4B, and 14B respectively, with consistent pass@k gains across all three scales.
  arXiv  Detail & Related papers  (2026-06-02T06:51:15Z)
• Off-the-Shelf LLMs as Process Scorers: Training-Free Alternative to PRMs for Mathematical Reasoning [51.88950852117154]
  Chunk-Level Guided Generation is a training-free alternative that uses an off-the-shelf large language model as a process scorer.<n>We show that scoring variable-length reasoning steps with large-model likelihoods is unreliable due to a systematic length bias.<n>Chunk-Level Guided Generation produces substantially shorter reasoning traces than PRM guided search.
  arXiv  Detail & Related papers  (2026-06-01T04:43:36Z)
• FormInv: A Measurement Protocol for Semantic Invariance in Mathematical Reasoning Benchmarks [0.0]
  A paraphrase-quality audit of MathCheck detected 4 semantically incorrect paraphrases in 129 groups.<n>Removing them drops GPT-4o from rank 2 to rank 4 and elevates Claude Haiku and DeepSeek V3 above it.
  arXiv  Detail & Related papers  (2026-05-27T18:59:18Z)
• 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)
• How Far Is Document Parsing from Solved? PureDocBench: A Source-TraceableBenchmark across Clean, Degraded, and Real-World Settings [56.70440596502351]
  The past year has seen over 20 open document parsing models, yet benchmarks almost exclusively on OmniDocBench.<n>We present PureDocBench, a benchmark thatrenders document images from HTML/CSS, covering 10 domains, 66 pages, and 1,475 pages.
  arXiv  Detail & Related papers  (2026-05-08T09:30:31Z)
• vla-eval: A Unified Evaluation Harness for Vision-Language-Action Models [58.633451339058986]
  VLA models are typically evaluated using per benchmark scripts maintained independently by each model repository.<n>We present vla eval, an open source evaluation harness that decouples model inference from benchmark execution.<n>A complete evaluation requires only two commands: vla eval serve and vla eval run.
  arXiv  Detail & Related papers  (2026-03-14T14:38:53Z)
• Preference Optimization for Reasoning with Pseudo Feedback [100.62603571434167]
  We introduce a novel approach to generate pseudo feedback for reasoning tasks by framing the labeling of solutions as an evaluation against associated test cases.<n>We conduct experiments on both mathematical reasoning and coding tasks using pseudo feedback for preference optimization, and observe improvements across both tasks.
  arXiv  Detail & Related papers  (2024-11-25T12:44:02Z)

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.