Anticipatory Evaluation of Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Park, Jungsoo, Mendes, Ethan, Stanovsky, Gabriel, Ritter, Alan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908812466716672
author Park, Jungsoo
Mendes, Ethan
Stanovsky, Gabriel
Ritter, Alan
author_facet Park, Jungsoo
Mendes, Ethan
Stanovsky, Gabriel
Ritter, Alan
contents Progress in large language models is increasingly constrained by an evaluation bottleneck: benchmarks must be built and models run before iteration can begin. We investigate whether evaluation outcomes can be forecast before any experiments are conducted. Specifically, we study text-only performance prediction, where models estimate performance from task descriptions and experimental configurations alone, without access to dataset instances. To support systematic study, we curate PRECOG, a corpus of description-performance pairs spanning diverse tasks, domains, and metrics. We scrape task and configuration descriptions from arXiv, yielding 2,290 instances covering 1,519 papers, and construct a test split using papers published after the evaluated models' knowledge cutoff. Experiments show the task is challenging but feasible: reasoning models achieve a non-trivial forecasting skill reaching mean absolute error as low as 9.9 at high-confidence thresholds. Overall, our corpus and analyses offer an initial step toward open-ended anticipatory evaluation, supporting difficulty estimation and smarter resource allocation.
format Preprint
id arxiv_https___arxiv_org_abs_2509_20645
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Anticipatory Evaluation of Language Models
Park, Jungsoo
Mendes, Ethan
Stanovsky, Gabriel
Ritter, Alan
Computation and Language
Artificial Intelligence
Machine Learning
Progress in large language models is increasingly constrained by an evaluation bottleneck: benchmarks must be built and models run before iteration can begin. We investigate whether evaluation outcomes can be forecast before any experiments are conducted. Specifically, we study text-only performance prediction, where models estimate performance from task descriptions and experimental configurations alone, without access to dataset instances. To support systematic study, we curate PRECOG, a corpus of description-performance pairs spanning diverse tasks, domains, and metrics. We scrape task and configuration descriptions from arXiv, yielding 2,290 instances covering 1,519 papers, and construct a test split using papers published after the evaluated models' knowledge cutoff. Experiments show the task is challenging but feasible: reasoning models achieve a non-trivial forecasting skill reaching mean absolute error as low as 9.9 at high-confidence thresholds. Overall, our corpus and analyses offer an initial step toward open-ended anticipatory evaluation, supporting difficulty estimation and smarter resource allocation.
title Anticipatory Evaluation of Language Models
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2509.20645