@inproceedings{biswas-etal-2026-prism,
title = "{PR}i{SM}: Partial Ranking via Inter-layer Semantic Measurement for Efficient Fine-tuning of Language Models",
author = "Biswas, Aldrin Kabya and
Fahim, Md and
Amin, Md. Ashraful and
Ali, Amin Ahsan and
Rahman, AKM Mahbubur",
editor = "Piperidis, Stelios and
Bel, N{\'u}ria and
van den Heuvel, Henk and
Ide, Nancy and
Krek, Simon and
Toral, Antonio",
booktitle = "Proceedings of the Fifteenth Language Resources and Evaluation Conference",
month = may,
year = "2026",
address = "Palma de Mallorca, Spain",
publisher = "ELRA Language Resource Association",
url = "https://aclanthology.org/2026.lrec-1.810/",
doi = "10.63317/3eyz8rr5qun6",
pages = "10313--10323",
abstract = "The growing scale of pre-trained language models poses a challenge in fine-tuning for downstream tasks, especially in resource-constrained settings. Recent studies highlight that not all layers in transformer-based language models contribute equally to downstream task performance, giving rise to various partial fine-tuning strategies. However, current methods often introduce significant training overhead or rely on simple heuristics that yield suboptimal performance and poor generalization. We propose PRiSM (Partial Ranking via inter-layer Semantic Measurement), a training-free approach for layer-wise partial fine-tuning that leverages the cosine similarity between pre-trained aggregate token representations across layers to identify inter-layer relationships. comprises two stages: (i) scoring layers based on their relevance to the task via a single forward pass, and (ii) fine-tuning a subset of block-wise highest-scoring layers, while keeping others frozen. We conduct experiments on 15 diverse NLP datasets, including single-sentence and sentence-pair classification tasks. Our method achieves competitive performance compared to full fine-tuning, with an average training speedup of 1.5{\texttimes} and a reduction of trainable parameters by 75{\%}, and outperforms all the comparative baselines. Additionally, our approach does not cause any notable drop in performance when the domain is changed for the evaluation tasks, demonstrating robust cross-domain generalizability."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="biswas-etal-2026-prism">
<titleInfo>
<title>PRiSM: Partial Ranking via Inter-layer Semantic Measurement for Efficient Fine-tuning of Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Aldrin</namePart>
<namePart type="given">Kabya</namePart>
<namePart type="family">Biswas</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Md</namePart>
<namePart type="family">Fahim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Md.</namePart>
<namePart type="given">Ashraful</namePart>
<namePart type="family">Amin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Amin</namePart>
<namePart type="given">Ahsan</namePart>
<namePart type="family">Ali</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">AKM</namePart>
<namePart type="given">Mahbubur</namePart>
<namePart type="family">Rahman</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Fifteenth Language Resources and Evaluation Conference</title>
</titleInfo>
<name type="personal">
<namePart type="given">Stelios</namePart>
<namePart type="family">Piperidis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Núria</namePart>
<namePart type="family">Bel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Henk</namePart>
<namePart type="family">van den Heuvel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nancy</namePart>
<namePart type="family">Ide</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Simon</namePart>
<namePart type="family">Krek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Antonio</namePart>
<namePart type="family">Toral</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resource Association</publisher>
<place>
<placeTerm type="text">Palma de Mallorca, Spain</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>The growing scale of pre-trained language models poses a challenge in fine-tuning for downstream tasks, especially in resource-constrained settings. Recent studies highlight that not all layers in transformer-based language models contribute equally to downstream task performance, giving rise to various partial fine-tuning strategies. However, current methods often introduce significant training overhead or rely on simple heuristics that yield suboptimal performance and poor generalization. We propose PRiSM (Partial Ranking via inter-layer Semantic Measurement), a training-free approach for layer-wise partial fine-tuning that leverages the cosine similarity between pre-trained aggregate token representations across layers to identify inter-layer relationships. comprises two stages: (i) scoring layers based on their relevance to the task via a single forward pass, and (ii) fine-tuning a subset of block-wise highest-scoring layers, while keeping others frozen. We conduct experiments on 15 diverse NLP datasets, including single-sentence and sentence-pair classification tasks. Our method achieves competitive performance compared to full fine-tuning, with an average training speedup of 1.5× and a reduction of trainable parameters by 75%, and outperforms all the comparative baselines. Additionally, our approach does not cause any notable drop in performance when the domain is changed for the evaluation tasks, demonstrating robust cross-domain generalizability.</abstract>
<identifier type="citekey">biswas-etal-2026-prism</identifier>
<identifier type="doi">10.63317/3eyz8rr5qun6</identifier>
<location>
<url>https://aclanthology.org/2026.lrec-1.810/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>10313</start>
<end>10323</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T PRiSM: Partial Ranking via Inter-layer Semantic Measurement for Efficient Fine-tuning of Language Models
%A Biswas, Aldrin Kabya
%A Fahim, Md
%A Amin, Md. Ashraful
%A Ali, Amin Ahsan
%A Rahman, AKM Mahbubur
%Y Piperidis, Stelios
%Y Bel, Núria
%Y van den Heuvel, Henk
%Y Ide, Nancy
%Y Krek, Simon
%Y Toral, Antonio
%S Proceedings of the Fifteenth Language Resources and Evaluation Conference
%D 2026
%8 May
%I ELRA Language Resource Association
%C Palma de Mallorca, Spain
%F biswas-etal-2026-prism
%X The growing scale of pre-trained language models poses a challenge in fine-tuning for downstream tasks, especially in resource-constrained settings. Recent studies highlight that not all layers in transformer-based language models contribute equally to downstream task performance, giving rise to various partial fine-tuning strategies. However, current methods often introduce significant training overhead or rely on simple heuristics that yield suboptimal performance and poor generalization. We propose PRiSM (Partial Ranking via inter-layer Semantic Measurement), a training-free approach for layer-wise partial fine-tuning that leverages the cosine similarity between pre-trained aggregate token representations across layers to identify inter-layer relationships. comprises two stages: (i) scoring layers based on their relevance to the task via a single forward pass, and (ii) fine-tuning a subset of block-wise highest-scoring layers, while keeping others frozen. We conduct experiments on 15 diverse NLP datasets, including single-sentence and sentence-pair classification tasks. Our method achieves competitive performance compared to full fine-tuning, with an average training speedup of 1.5× and a reduction of trainable parameters by 75%, and outperforms all the comparative baselines. Additionally, our approach does not cause any notable drop in performance when the domain is changed for the evaluation tasks, demonstrating robust cross-domain generalizability.
%R 10.63317/3eyz8rr5qun6
%U https://aclanthology.org/2026.lrec-1.810/
%U https://doi.org/10.63317/3eyz8rr5qun6
%P 10313-10323
Markdown (Informal)
[PRiSM: Partial Ranking via Inter-layer Semantic Measurement for Efficient Fine-tuning of Language Models](https://aclanthology.org/2026.lrec-1.810/) (Biswas et al., LREC 2026)
ACL