@inproceedings{yang-etal-2026-rocob,
title = "{ROCOB}: Dual-Manifold Alignment with Optimal Budgeting for Training-Free {PTQ} Recovery",
author = "Yang, Haoqi and
Xiao, Teng and
Li, Zuchao and
Huang, Hao and
Zhang, Lefei and
Zhao, Hai",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the {A}sia-Pacific Chapter of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.aacl-long.158/",
pages = "3491--3512",
ISBN = "979-8-89176-439-2",
abstract = "Post-training quantization facilitates efficient LLM deployment, yet it disproportionately erodes reasoning and mathematical performance. To recover fidelity without costly fine-tuning, we identify a critical vulnerability in existing output-space methods: the tendency to overfit the limited calibration manifold, which necessitates structural regularization for unseen data. We propose **ROCOB** (**R**esilient **O**utput **C**ompensation with **O**ptimal **B**udgeting), a training-free framework based on *Dual-Manifold Alignment*. ROCOB couples a Principal Activation Component (PAC), which minimizes error on the calibration manifold, with an Orthogonal Geometric Component (OGC), which recovers residual weight-space structure without fitting a second calibration objective. To compare methods fairly under the same added-parameter budget, ROCOB uses Global Spectral Energy Maximization (GSEM) to allocate ranks according to recovered spectral energy and module cost. Across Qwen3 and Llama-3 models, quantizers, and reasoning benchmarks, ROCOB improves average post-quantization recovery in most evaluated model{--}quantizer settings while preserving a training-free construction pipeline."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yang-etal-2026-rocob">
<titleInfo>
<title>ROCOB: Dual-Manifold Alignment with Optimal Budgeting for Training-Free PTQ Recovery</title>
</titleInfo>
<name type="personal">
<namePart type="given">Haoqi</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Teng</namePart>
<namePart type="family">Xiao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zuchao</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hao</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lefei</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hai</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-439-2</identifier>
</relatedItem>
<abstract>Post-training quantization facilitates efficient LLM deployment, yet it disproportionately erodes reasoning and mathematical performance. To recover fidelity without costly fine-tuning, we identify a critical vulnerability in existing output-space methods: the tendency to overfit the limited calibration manifold, which necessitates structural regularization for unseen data. We propose **ROCOB** (**R**esilient **O**utput **C**ompensation with **O**ptimal **B**udgeting), a training-free framework based on *Dual-Manifold Alignment*. ROCOB couples a Principal Activation Component (PAC), which minimizes error on the calibration manifold, with an Orthogonal Geometric Component (OGC), which recovers residual weight-space structure without fitting a second calibration objective. To compare methods fairly under the same added-parameter budget, ROCOB uses Global Spectral Energy Maximization (GSEM) to allocate ranks according to recovered spectral energy and module cost. Across Qwen3 and Llama-3 models, quantizers, and reasoning benchmarks, ROCOB improves average post-quantization recovery in most evaluated model–quantizer settings while preserving a training-free construction pipeline.</abstract>
<identifier type="citekey">yang-etal-2026-rocob</identifier>
<location>
<url>https://aclanthology.org/2026.aacl-long.158/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>3491</start>
<end>3512</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T ROCOB: Dual-Manifold Alignment with Optimal Budgeting for Training-Free PTQ Recovery
%A Yang, Haoqi
%A Xiao, Teng
%A Li, Zuchao
%A Huang, Hao
%A Zhang, Lefei
%A Zhao, Hai
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-439-2
%F yang-etal-2026-rocob
%X Post-training quantization facilitates efficient LLM deployment, yet it disproportionately erodes reasoning and mathematical performance. To recover fidelity without costly fine-tuning, we identify a critical vulnerability in existing output-space methods: the tendency to overfit the limited calibration manifold, which necessitates structural regularization for unseen data. We propose **ROCOB** (**R**esilient **O**utput **C**ompensation with **O**ptimal **B**udgeting), a training-free framework based on *Dual-Manifold Alignment*. ROCOB couples a Principal Activation Component (PAC), which minimizes error on the calibration manifold, with an Orthogonal Geometric Component (OGC), which recovers residual weight-space structure without fitting a second calibration objective. To compare methods fairly under the same added-parameter budget, ROCOB uses Global Spectral Energy Maximization (GSEM) to allocate ranks according to recovered spectral energy and module cost. Across Qwen3 and Llama-3 models, quantizers, and reasoning benchmarks, ROCOB improves average post-quantization recovery in most evaluated model–quantizer settings while preserving a training-free construction pipeline.
%U https://aclanthology.org/2026.aacl-long.158/
%P 3491-3512
Markdown (Informal)
[ROCOB: Dual-Manifold Alignment with Optimal Budgeting for Training-Free PTQ Recovery](https://aclanthology.org/2026.aacl-long.158/) (Yang et al., AACL-IJCNLP 2026)
ACL
- Haoqi Yang, Teng Xiao, Zuchao Li, Hao Huang, Lefei Zhang, and Hai Zhao. 2026. ROCOB: Dual-Manifold Alignment with Optimal Budgeting for Training-Free PTQ Recovery. In Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3491–3512, Hengqin, China. Association for Computational Linguistics.