@inproceedings{mule-etal-2026-teaching,
title = "Teaching Language Models to Forecast Research Success Through Comparative Idea Evaluation",
author = "Mule, Srujan P and
Garikaparthi, Aniketh and
Patwardhan, Manasi",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-acl.1918/",
doi = "10.18653/v1/2026.findings-acl.1918",
pages = "38491--38529",
ISBN = "979-8-89176-395-1",
abstract = "As language models accelerate scientific research by automating hypothesis generation and implementation, a new bottleneck emerges: evaluating and filtering hundreds of AI-generated ideas without exhaustive experimentation. We ask whether LMs can learn to forecast the empirical success of research ideas before any experiments are run. We study \textit{comparative empirical forecasting}: given a benchmark-specific research goal and two candidate ideas, predict which will achieve better benchmark performance. We construct a dataset of 11,488 idea pairs grounded in objective outcomes from PapersWithCode. While off-the-shelf 8B-parameter models struggle (30{\%} acc.), SFT dramatically boosts performance to 77.1{\%}, outperforming GPT-5 (61.1{\%}). By framing evaluation as a reasoning task via Reinforcement Learning with Verifiable Rewards (RLVR), we train models to discover latent reasoning paths, achieving 71.35{\%} acc. with interpretable justifications. Through additional ablations and out-of-distribution tests, we show robustness to surface-level heuristics and transfer to both a cross-domain time-split test set and an independently constructed test set. Our results demonstrate that compute-efficient small language models can serve as effective, objective verifiers, offering a scalable path for autonomous scientific discovery."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="mule-etal-2026-teaching">
<titleInfo>
<title>Teaching Language Models to Forecast Research Success Through Comparative Idea Evaluation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Srujan</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Mule</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aniketh</namePart>
<namePart type="family">Garikaparthi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Manasi</namePart>
<namePart type="family">Patwardhan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-395-1</identifier>
</relatedItem>
<abstract>As language models accelerate scientific research by automating hypothesis generation and implementation, a new bottleneck emerges: evaluating and filtering hundreds of AI-generated ideas without exhaustive experimentation. We ask whether LMs can learn to forecast the empirical success of research ideas before any experiments are run. We study comparative empirical forecasting: given a benchmark-specific research goal and two candidate ideas, predict which will achieve better benchmark performance. We construct a dataset of 11,488 idea pairs grounded in objective outcomes from PapersWithCode. While off-the-shelf 8B-parameter models struggle (30% acc.), SFT dramatically boosts performance to 77.1%, outperforming GPT-5 (61.1%). By framing evaluation as a reasoning task via Reinforcement Learning with Verifiable Rewards (RLVR), we train models to discover latent reasoning paths, achieving 71.35% acc. with interpretable justifications. Through additional ablations and out-of-distribution tests, we show robustness to surface-level heuristics and transfer to both a cross-domain time-split test set and an independently constructed test set. Our results demonstrate that compute-efficient small language models can serve as effective, objective verifiers, offering a scalable path for autonomous scientific discovery.</abstract>
<identifier type="citekey">mule-etal-2026-teaching</identifier>
<identifier type="doi">10.18653/v1/2026.findings-acl.1918</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.1918/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>38491</start>
<end>38529</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Teaching Language Models to Forecast Research Success Through Comparative Idea Evaluation
%A Mule, Srujan P.
%A Garikaparthi, Aniketh
%A Patwardhan, Manasi
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F mule-etal-2026-teaching
%X As language models accelerate scientific research by automating hypothesis generation and implementation, a new bottleneck emerges: evaluating and filtering hundreds of AI-generated ideas without exhaustive experimentation. We ask whether LMs can learn to forecast the empirical success of research ideas before any experiments are run. We study comparative empirical forecasting: given a benchmark-specific research goal and two candidate ideas, predict which will achieve better benchmark performance. We construct a dataset of 11,488 idea pairs grounded in objective outcomes from PapersWithCode. While off-the-shelf 8B-parameter models struggle (30% acc.), SFT dramatically boosts performance to 77.1%, outperforming GPT-5 (61.1%). By framing evaluation as a reasoning task via Reinforcement Learning with Verifiable Rewards (RLVR), we train models to discover latent reasoning paths, achieving 71.35% acc. with interpretable justifications. Through additional ablations and out-of-distribution tests, we show robustness to surface-level heuristics and transfer to both a cross-domain time-split test set and an independently constructed test set. Our results demonstrate that compute-efficient small language models can serve as effective, objective verifiers, offering a scalable path for autonomous scientific discovery.
%R 10.18653/v1/2026.findings-acl.1918
%U https://aclanthology.org/2026.findings-acl.1918/
%U https://doi.org/10.18653/v1/2026.findings-acl.1918
%P 38491-38529
Markdown (Informal)
[Teaching Language Models to Forecast Research Success Through Comparative Idea Evaluation](https://aclanthology.org/2026.findings-acl.1918/) (Mule et al., Findings 2026)
ACL