@inproceedings{shah-etal-2026-trajectory,
title = "Trajectory Self-Feedback in Diffusion Language Models: A Systematic Survey",
author = "Shah, Yash and
Chakraborty, Abhijit and
Gupta, Vivek",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.1739/",
doi = "10.18653/2026.findings-emnlp.1739",
pages = "37189--37208",
ISBN = "979-8-89176-462-0",
abstract = "Diffusion language models generate text through iterative denoising, allowing intermediate predictions to influence later denoising steps, a recurring practice we call trajectory self-feedback (TSF). TSF is adopted under inconsistent terminology, evaluated on incompatible benchmarks, and studied by disconnected communities, yet no prior review characterises it reproducibly. We present a systematic literature review under an evidence-based methodology: six pre-specified questions, a documented six-library search with snowballing, explicit selection and quality criteria, and a structured synthesis of 24 primary studies across seven clusters, with a released audit trail. We find: (i) a model-agnostic taxonomy resolves four operator families: Prediction-Space Concatenation, Confidence-weighted Path Signals, Reward-Reinforced Conditioning, and Correction-Targeted Feedback, that the literature conflates as self-conditioning; (ii) evidence of benefit is real but fragmented across incomparable settings; (iii) within our search window, no study jointly varies the TSF mechanism and measures faithfulness on knowledge-intensive tasks; and (iv) a discrete/continuous asymmetry consistent with the sampling-wall hypothesis is visible but never tested. We conclude with eight open problems, a dependency map, and review-derived diagnostics (Commitment Index, Correction Capacity) and a Standardised Evaluation Protocol (SEP-DLM) to make TSF measurable across quality, correction, trajectory commitment, and faithfulness."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="shah-etal-2026-trajectory">
<titleInfo>
<title>Trajectory Self-Feedback in Diffusion Language Models: A Systematic Survey</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yash</namePart>
<namePart type="family">Shah</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Abhijit</namePart>
<namePart type="family">Chakraborty</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vivek</namePart>
<namePart type="family">Gupta</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>Findings of the Association for Computational Linguistics: EMNLP 2026</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">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Diffusion language models generate text through iterative denoising, allowing intermediate predictions to influence later denoising steps, a recurring practice we call trajectory self-feedback (TSF). TSF is adopted under inconsistent terminology, evaluated on incompatible benchmarks, and studied by disconnected communities, yet no prior review characterises it reproducibly. We present a systematic literature review under an evidence-based methodology: six pre-specified questions, a documented six-library search with snowballing, explicit selection and quality criteria, and a structured synthesis of 24 primary studies across seven clusters, with a released audit trail. We find: (i) a model-agnostic taxonomy resolves four operator families: Prediction-Space Concatenation, Confidence-weighted Path Signals, Reward-Reinforced Conditioning, and Correction-Targeted Feedback, that the literature conflates as self-conditioning; (ii) evidence of benefit is real but fragmented across incomparable settings; (iii) within our search window, no study jointly varies the TSF mechanism and measures faithfulness on knowledge-intensive tasks; and (iv) a discrete/continuous asymmetry consistent with the sampling-wall hypothesis is visible but never tested. We conclude with eight open problems, a dependency map, and review-derived diagnostics (Commitment Index, Correction Capacity) and a Standardised Evaluation Protocol (SEP-DLM) to make TSF measurable across quality, correction, trajectory commitment, and faithfulness.</abstract>
<identifier type="citekey">shah-etal-2026-trajectory</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.1739</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1739/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>37189</start>
<end>37208</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Trajectory Self-Feedback in Diffusion Language Models: A Systematic Survey
%A Shah, Yash
%A Chakraborty, Abhijit
%A Gupta, Vivek
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F shah-etal-2026-trajectory
%X Diffusion language models generate text through iterative denoising, allowing intermediate predictions to influence later denoising steps, a recurring practice we call trajectory self-feedback (TSF). TSF is adopted under inconsistent terminology, evaluated on incompatible benchmarks, and studied by disconnected communities, yet no prior review characterises it reproducibly. We present a systematic literature review under an evidence-based methodology: six pre-specified questions, a documented six-library search with snowballing, explicit selection and quality criteria, and a structured synthesis of 24 primary studies across seven clusters, with a released audit trail. We find: (i) a model-agnostic taxonomy resolves four operator families: Prediction-Space Concatenation, Confidence-weighted Path Signals, Reward-Reinforced Conditioning, and Correction-Targeted Feedback, that the literature conflates as self-conditioning; (ii) evidence of benefit is real but fragmented across incomparable settings; (iii) within our search window, no study jointly varies the TSF mechanism and measures faithfulness on knowledge-intensive tasks; and (iv) a discrete/continuous asymmetry consistent with the sampling-wall hypothesis is visible but never tested. We conclude with eight open problems, a dependency map, and review-derived diagnostics (Commitment Index, Correction Capacity) and a Standardised Evaluation Protocol (SEP-DLM) to make TSF measurable across quality, correction, trajectory commitment, and faithfulness.
%R 10.18653/2026.findings-emnlp.1739
%U https://aclanthology.org/2026.findings-emnlp.1739/
%U https://doi.org/10.18653/2026.findings-emnlp.1739
%P 37189-37208
Markdown (Informal)
[Trajectory Self-Feedback in Diffusion Language Models: A Systematic Survey](https://aclanthology.org/2026.findings-emnlp.1739/) (Shah et al., Findings 2026)
ACL