@inproceedings{charkiewicz-nowakowski-2026-laniqo,
title = "Laniqo at {WMT}26 Multilingual Instruction Shared Task ({MIST}): Full-Parameter Knowledge Distillation with Task-Conditional Pivot Decoding Under a 10{B}-Parameter Constraint",
author = "Charkiewicz, Adrian and
Nowakowski, Artur",
editor = "Haddow, Barry and
Kocmi, Tom and
Koehn, Philipp and
Monz, Christof",
booktitle = "Proceedings of the Eleventh Conference on Machine Translation",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.wmt-main.126/",
pages = "1906--1916",
ISBN = "979-8-89176-474-3",
abstract = "We describe our submission to the WMT26 Multilingual Instruction Shared Task (MIST). Our system is google/gemma-4-E4B-it (8B total / 4.5B effective parameters), fully fine-tuned via knowledge distillation from a larger teacher model (Gemma-4-31B-it) on teacher-generated responses spanning three sub-tasks and 24 languages. We find that (i) a two-judge quality-agreement filter on the distillation data provides no measurable benefit over using the teacher{'}s outputs unfiltered, at both LoRA and full-parameter training scale, and (ii) a task- conditional decoding strategy (using two-turn English drafting for open-ended generation and summarization, versus single-turn direct decod- ing for context-grounded question answering) significantly improves quality where applied and is harmless where it is not. Our final submission combines full-parameter distillation with this task-conditional decoding scheme."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="charkiewicz-nowakowski-2026-laniqo">
<titleInfo>
<title>Laniqo at WMT26 Multilingual Instruction Shared Task (MIST): Full-Parameter Knowledge Distillation with Task-Conditional Pivot Decoding Under a 10B-Parameter Constraint</title>
</titleInfo>
<name type="personal">
<namePart type="given">Adrian</namePart>
<namePart type="family">Charkiewicz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Artur</namePart>
<namePart type="family">Nowakowski</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 Eleventh Conference on Machine Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Barry</namePart>
<namePart type="family">Haddow</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tom</namePart>
<namePart type="family">Kocmi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Philipp</namePart>
<namePart type="family">Koehn</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Christof</namePart>
<namePart type="family">Monz</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">979-8-89176-474-3</identifier>
</relatedItem>
<abstract>We describe our submission to the WMT26 Multilingual Instruction Shared Task (MIST). Our system is google/gemma-4-E4B-it (8B total / 4.5B effective parameters), fully fine-tuned via knowledge distillation from a larger teacher model (Gemma-4-31B-it) on teacher-generated responses spanning three sub-tasks and 24 languages. We find that (i) a two-judge quality-agreement filter on the distillation data provides no measurable benefit over using the teacher’s outputs unfiltered, at both LoRA and full-parameter training scale, and (ii) a task- conditional decoding strategy (using two-turn English drafting for open-ended generation and summarization, versus single-turn direct decod- ing for context-grounded question answering) significantly improves quality where applied and is harmless where it is not. Our final submission combines full-parameter distillation with this task-conditional decoding scheme.</abstract>
<identifier type="citekey">charkiewicz-nowakowski-2026-laniqo</identifier>
<location>
<url>https://aclanthology.org/2026.wmt-main.126/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>1906</start>
<end>1916</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Laniqo at WMT26 Multilingual Instruction Shared Task (MIST): Full-Parameter Knowledge Distillation with Task-Conditional Pivot Decoding Under a 10B-Parameter Constraint
%A Charkiewicz, Adrian
%A Nowakowski, Artur
%Y Haddow, Barry
%Y Kocmi, Tom
%Y Koehn, Philipp
%Y Monz, Christof
%S Proceedings of the Eleventh Conference on Machine Translation
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-474-3
%F charkiewicz-nowakowski-2026-laniqo
%X We describe our submission to the WMT26 Multilingual Instruction Shared Task (MIST). Our system is google/gemma-4-E4B-it (8B total / 4.5B effective parameters), fully fine-tuned via knowledge distillation from a larger teacher model (Gemma-4-31B-it) on teacher-generated responses spanning three sub-tasks and 24 languages. We find that (i) a two-judge quality-agreement filter on the distillation data provides no measurable benefit over using the teacher’s outputs unfiltered, at both LoRA and full-parameter training scale, and (ii) a task- conditional decoding strategy (using two-turn English drafting for open-ended generation and summarization, versus single-turn direct decod- ing for context-grounded question answering) significantly improves quality where applied and is harmless where it is not. Our final submission combines full-parameter distillation with this task-conditional decoding scheme.
%U https://aclanthology.org/2026.wmt-main.126/
%P 1906-1916
Markdown (Informal)
[Laniqo at WMT26 Multilingual Instruction Shared Task (MIST): Full-Parameter Knowledge Distillation with Task-Conditional Pivot Decoding Under a 10B-Parameter Constraint](https://aclanthology.org/2026.wmt-main.126/) (Charkiewicz & Nowakowski, WMT 2026)
ACL