Train-before-Test Harmonizes Language Model Rankings

• URL: http://arxiv.org/abs/2507.05195v2
• Date: Mon, 13 Oct 2025 12:21:52 GMT
• Title: Train-before-Test Harmonizes Language Model Rankings
• Abstract summary: Existing language model benchmarks provide contradictory model rankings, even for benchmarks that aim to capture similar skills.<n>We compare model potential by providing each model with identical benchmark-specific fine-tuning before evaluation.<n>We show that model potential rankings obtained through train-before-test exhibit remarkable consistency across all benchmarks.
• Score: 39.02785699830391
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Existing language model benchmarks provide contradictory model rankings, even for benchmarks that aim to capture similar skills. This dilemma of conflicting rankings hampers model selection, clouds model comparisons, and adds confusion to a growing ecosystem of competing models. In this paper, we take a different perspective on model comparison: instead of relying on out-of-the-box performance via direct evaluation, we compare model potential by providing each model with identical benchmark-specific fine-tuning before evaluation. We call this approach train-before-test. Our primary contribution is a comprehensive empirical evaluation of model potential across 24 benchmarks and 61 models. First, we demonstrate that model potential rankings obtained through train-before-test exhibit remarkable consistency across all benchmarks. Whereas traditional rankings demonstrate little external validity under direct evaluation, they enjoy a significant degree of external validity when applying train-before-test: model potential rankings transfer gracefully from one benchmark to another. Second, train-before-test restores the connection between perplexity and downstream task performance, lost under direct evaluation. Remarkably, even pre-finetuning perplexity of a base model predicts post-finetuning downstream performance, suggesting that ranking consistency reflects inherent model potential rather than fine-tuning artifacts. Finally, train-before-test reduces the model-score matrix to essentially rank one, indicating that model potential is dominated by one latent factor, uncovered by train-before-test. Our work supports the recommendation to make train-before-test a default component of LLM benchmarking.

Related papers

• Impression Share Prediction: An Offline Evaluation Task for Ranking Systems [2.455009405787869]
  We propose emphimpression share prediction as an offline evaluation task.<n>The task is inherently counterfactual, since the candidate has never served live traffic.<n>We develop a statistical learning framework that predicts impression shares from a candidate's early-interaction confidence signals.
  arXiv  Detail & Related papers  (2026-08-17T17:52:55Z)
• AI4AI at Test-Time: Strong-to-Weak Capability Transfer via Harnesses [72.0777620499307]
  We study whether a stronger builder model can construct inference-time harnesses that help a weaker target model solve tasks more reliably without any parameter updates.<n> Empirically, this form of test-time capability transfer is highly effective, nearly doubling average target-model performance from 0.49 to 0.91.<n>We find that builder-model reasoning effort improves harness quality monotonically, platform effects are modest relative to the builder model's own capability, and weaker target models receive the largest gains.
  arXiv  Detail & Related papers  (2026-08-12T17:53:18Z)
• On Predicting the Post-training Potential of Pre-trained LLMs [60.07459271263409]
  We introduce a new task of predicting post-training potential - forecasting a base model's performance before post-training.<n>We propose RuDE, a unified framework that bypasses the generation gap of base models by leveraging response discrimination.<n>Experiments demonstrate a correlation greater than 90% with post-training performance.
  arXiv  Detail & Related papers  (2026-05-12T11:33:49Z)
• Test-time RL alignment exposes task familiarity artifacts in LLM benchmarks [20.988724384459392]
  The train-before-test approach controls for task familiarity by giving each model task-relevant training before evaluation.<n>We propose a two-stage test-time reinforcement learning (RL) alignment method for train-before-test.
  arXiv  Detail & Related papers  (2026-03-13T10:24:19Z)
• RewardBench 2: Advancing Reward Model Evaluation [71.65938693914153]
  Reward models are used throughout the post-training of language models to capture nuanced signals from preference data.<n>The community has begun establishing best practices for evaluating reward models.<n>This paper introduces RewardBench 2, a new multi-skill reward modeling benchmark.
  arXiv  Detail & Related papers  (2025-06-02T17:54:04Z)
• Where is this coming from? Making groundedness count in the evaluation of Document VQA models [12.951716701565019]
  We argue that common evaluation metrics do not account for the semantic and multimodal groundedness of a model's outputs.<n>We propose a new evaluation methodology that accounts for the groundedness of predictions.<n>Our proposed methodology is parameterized in such a way that users can configure the score according to their preferences.
  arXiv  Detail & Related papers  (2025-03-24T20:14:46Z)
• Prior2Posterior: Model Prior Correction for Long-Tailed Learning [0.41248472494152805]
  We propose a novel approach to accurately model the effective prior of a trained model using textita posteriori probabilities.<n>We show that the proposed approach achieves new state-of-the-art (SOTA) on several benchmark datasets from the long-tail literature.
  arXiv  Detail & Related papers  (2024-12-21T08:49:02Z)
• Self-Taught Evaluators [77.92610887220594]
  We present an approach that aims to im-proves without human annotations, using synthetic training data only. Our Self-Taught Evaluator can improve a strong LLM from 75.4 to 88.3 on RewardBench.
  arXiv  Detail & Related papers  (2024-08-05T17:57:02Z)
• Importance of Disjoint Sampling in Conventional and Transformer Models for Hyperspectral Image Classification [2.1223532600703385]
  This paper presents an innovative disjoint sampling approach for training SOTA models on Hyperspectral image classification (HSIC) tasks. By separating training, validation, and test data without overlap, the proposed method facilitates a fairer evaluation of how well a model can classify pixels it was not exposed to during training or validation. This rigorous methodology is critical for advancing SOTA models and their real-world application to large-scale land mapping with Hyperspectral sensors.
  arXiv  Detail & Related papers  (2024-04-23T11:40:52Z)
• Don't Make Your LLM an Evaluation Benchmark Cheater [142.24553056600627]
  Large language models(LLMs) have greatly advanced the frontiers of artificial intelligence, attaining remarkable improvement in model capacity. To assess the model performance, a typical approach is to construct evaluation benchmarks for measuring the ability level of LLMs. We discuss the potential risk and impact of inappropriately using evaluation benchmarks and misleadingly interpreting the evaluation results.
  arXiv  Detail & Related papers  (2023-11-03T14:59:54Z)
• How to Estimate Model Transferability of Pre-Trained Speech Models? [84.11085139766108]
  "Score-based assessment" framework for estimating transferability of pre-trained speech models. We leverage upon two representation theories, Bayesian likelihood estimation and optimal transport, to generate rank scores for the PSM candidates. Our framework efficiently computes transferability scores without actual fine-tuning of candidate models or layers.
  arXiv  Detail & Related papers  (2023-06-01T04:52:26Z)
• Guide the Learner: Controlling Product of Experts Debiasing Method Based on Token Attribution Similarities [17.082695183953486]
  A popular workaround is to train a robust model by re-weighting training examples based on a secondary biased model. Here, the underlying assumption is that the biased model resorts to shortcut features. We introduce a fine-tuning strategy that incorporates the similarity between the main and biased model attribution scores in a Product of Experts loss function.
  arXiv  Detail & Related papers  (2023-02-06T15:21:41Z)
• Unifying Language Learning Paradigms [96.35981503087567]
  We present a unified framework for pre-training models that are universally effective across datasets and setups. We show how different pre-training objectives can be cast as one another and how interpolating between different objectives can be effective. Our model also achieve strong results at in-context learning, outperforming 175B GPT-3 on zero-shot SuperGLUE and tripling the performance of T5-XXL on one-shot summarization.
  arXiv  Detail & Related papers  (2022-05-10T19:32:20Z)
• End-to-End Weak Supervision [15.125993628007972]
  We propose an end-to-end approach for directly learning the downstream model. We show improved performance over prior work in terms of end model performance on downstream test sets.
  arXiv  Detail & Related papers  (2021-07-05T19:10:11Z)
• Positive-Congruent Training: Towards Regression-Free Model Updates [87.25247195148187]
  In image classification, sample-wise inconsistencies appear as "negative flips" A new model incorrectly predicts the output for a test sample that was correctly classified by the old (reference) model. We propose a simple approach for PC training, Focal Distillation, which enforces congruence with the reference model.
  arXiv  Detail & Related papers  (2020-11-18T09:00:44Z)
• Monotonicity in practice of adaptive testing [0.0]
  This article evaluates Bayesian network models used for computerized adaptive testing and learned with a recently proposed monotonicity gradient algorithm. The quality of methods is empirically evaluated on a large data set of the Czech National Mathematics exam.
  arXiv  Detail & Related papers  (2020-09-15T10:55:41Z)
• Pre-training Is (Almost) All You Need: An Application to Commonsense Reasoning [61.32992639292889]
  Fine-tuning of pre-trained transformer models has become the standard approach for solving common NLP tasks. We introduce a new scoring method that casts a plausibility ranking task in a full-text format. We show that our method provides a much more stable training phase across random restarts.
  arXiv  Detail & Related papers  (2020-04-29T10:54:40Z)
• Document Ranking with a Pretrained Sequence-to-Sequence Model [56.44269917346376]
  We show how a sequence-to-sequence model can be trained to generate relevance labels as "target words" Our approach significantly outperforms an encoder-only model in a data-poor regime.
  arXiv  Detail & Related papers  (2020-03-14T22:29:50Z)

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.