@inproceedings{abdallah-etal-2026-smallreason,
title = "{S}mall{R}eason-{C}ol{BERT}: An Ultra-Small Late-Interaction Retriever for Reasoning Intensive Retrieval",
author = "Abdallah, Abdelrahman and
Ali, Mohammed and
Jatowt, Adam",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.460/",
pages = "10564--10575",
ISBN = "979-8-89176-459-0",
abstract = "Reasoning-intensive retrieval remains difficult for small models. Compact public ColBERTs are usually trained on general-purpose corpora and underperform reasoning-tuned 150M+ baselines on BRIGHT{~}(CITATION) by several nDCG@10 points; no public reasoning-tuned ColBERT exists at edge scale. We introduce \textbf{SmallReason-ColBERT}, a 32M late-interaction retriever that closes much of this gap with three ingredients: a varied-length contrastive warmup on ReasonIR-VL, a hard-negative distillation polish on merged ReasonIR-HQ and BGE-Reasoner data, and a single-layer per-query-token importance head trained on top of the frozen base. The head is trained with an un-normalised weighted MaxSim score and evaluated with its length-normalised form. In a controlled re-training, replacing this training objective with the symmetric normalised score causes the loss to stall and costs 3.59 nDCG@10. The full recipe reaches \textbf{21.41} mean nDCG@10 on BRIGHT, within 1.21 of the 150M Reason-ModernColBERT (22.62) and above all $\le 33$M ColBERTs we evaluate. Through ablations on capacity, initialisation, and score variants, we further show that the learned head outperforms fixed IDF weighting and that simply thresholding the learned gates is harmful."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="abdallah-etal-2026-smallreason">
<titleInfo>
<title>SmallReason-ColBERT: An Ultra-Small Late-Interaction Retriever for Reasoning Intensive Retrieval</title>
</titleInfo>
<name type="personal">
<namePart type="given">Abdelrahman</namePart>
<namePart type="family">Abdallah</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammed</namePart>
<namePart type="family">Ali</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Adam</namePart>
<namePart type="family">Jatowt</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Reasoning-intensive retrieval remains difficult for small models. Compact public ColBERTs are usually trained on general-purpose corpora and underperform reasoning-tuned 150M+ baselines on BRIGHT (CITATION) by several nDCG@10 points; no public reasoning-tuned ColBERT exists at edge scale. We introduce SmallReason-ColBERT, a 32M late-interaction retriever that closes much of this gap with three ingredients: a varied-length contrastive warmup on ReasonIR-VL, a hard-negative distillation polish on merged ReasonIR-HQ and BGE-Reasoner data, and a single-layer per-query-token importance head trained on top of the frozen base. The head is trained with an un-normalised weighted MaxSim score and evaluated with its length-normalised form. In a controlled re-training, replacing this training objective with the symmetric normalised score causes the loss to stall and costs 3.59 nDCG@10. The full recipe reaches 21.41 mean nDCG@10 on BRIGHT, within 1.21 of the 150M Reason-ModernColBERT (22.62) and above all łe 33M ColBERTs we evaluate. Through ablations on capacity, initialisation, and score variants, we further show that the learned head outperforms fixed IDF weighting and that simply thresholding the learned gates is harmful.</abstract>
<identifier type="citekey">abdallah-etal-2026-smallreason</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.460/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>10564</start>
<end>10575</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SmallReason-ColBERT: An Ultra-Small Late-Interaction Retriever for Reasoning Intensive Retrieval
%A Abdallah, Abdelrahman
%A Ali, Mohammed
%A Jatowt, Adam
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F abdallah-etal-2026-smallreason
%X Reasoning-intensive retrieval remains difficult for small models. Compact public ColBERTs are usually trained on general-purpose corpora and underperform reasoning-tuned 150M+ baselines on BRIGHT (CITATION) by several nDCG@10 points; no public reasoning-tuned ColBERT exists at edge scale. We introduce SmallReason-ColBERT, a 32M late-interaction retriever that closes much of this gap with three ingredients: a varied-length contrastive warmup on ReasonIR-VL, a hard-negative distillation polish on merged ReasonIR-HQ and BGE-Reasoner data, and a single-layer per-query-token importance head trained on top of the frozen base. The head is trained with an un-normalised weighted MaxSim score and evaluated with its length-normalised form. In a controlled re-training, replacing this training objective with the symmetric normalised score causes the loss to stall and costs 3.59 nDCG@10. The full recipe reaches 21.41 mean nDCG@10 on BRIGHT, within 1.21 of the 150M Reason-ModernColBERT (22.62) and above all łe 33M ColBERTs we evaluate. Through ablations on capacity, initialisation, and score variants, we further show that the learned head outperforms fixed IDF weighting and that simply thresholding the learned gates is harmful.
%U https://aclanthology.org/2026.emnlp-main.460/
%P 10564-10575
Markdown (Informal)
[SmallReason-ColBERT: An Ultra-Small Late-Interaction Retriever for Reasoning Intensive Retrieval](https://aclanthology.org/2026.emnlp-main.460/) (Abdallah et al., EMNLP 2026)
ACL