@inproceedings{jiang-etal-2026-imagine,
title = "Imagine Before You Predict: Interleaved Latent Visual Reasoning for Video Event Prediction",
author = "Jiang, Tianxiang and
Wu, Linquan and
Xia, Sheng and
Li, Songze and
Yan, Ziang and
Yang, Haoyu and
Qiao, Yu and
Wang, Yi",
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.2315/",
pages = "52600--52621",
ISBN = "979-8-89176-459-0",
abstract = "Video event prediction (VEP) requires models to infer unobserved future states from partial video evidence. Existing video MLLMs usually verbalize intermediate future reasoning in text space: once visual evidence is verbalized, fine-grained motion, geometry, and interaction cues can be lost, leading to plausible but visually ungrounded hallucinations. We introduce **Future-L1**, an **interleaved latent visual reasoning framework** that lets an MLLM alternate between language tokens and continuous latent visual spans during autoregressive decoding. To train this capability, we construct **Future-L1-50K** by selecting examples where future visual hints help prediction and aligning latent states to future-frame embeddings, then further optimizing sampled latent trajectories with **LA-DAPO**, a **latent-aware RL** objective with outcome-contrastive and temporal-diversity rewards. **Future-L1** achieves new **state-of-the-art** results on both benchmarks: on FutureBench, it improves Qwen3-VL-8B from 61.0 to **85.4** and exceeds the previous best Video-CoE by 10.4 points; on TwiFF-Bench, it improves the average score from 2.44 to **3.04**. These results suggest that future-oriented video reasoning benefits from preserving intermediate visual semantics in latent space rather than translating every reasoning step into text."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="jiang-etal-2026-imagine">
<titleInfo>
<title>Imagine Before You Predict: Interleaved Latent Visual Reasoning for Video Event Prediction</title>
</titleInfo>
<name type="personal">
<namePart type="given">Tianxiang</namePart>
<namePart type="family">Jiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Linquan</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sheng</namePart>
<namePart type="family">Xia</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Songze</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ziang</namePart>
<namePart type="family">Yan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haoyu</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yu</namePart>
<namePart type="family">Qiao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yi</namePart>
<namePart type="family">Wang</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>Video event prediction (VEP) requires models to infer unobserved future states from partial video evidence. Existing video MLLMs usually verbalize intermediate future reasoning in text space: once visual evidence is verbalized, fine-grained motion, geometry, and interaction cues can be lost, leading to plausible but visually ungrounded hallucinations. We introduce **Future-L1**, an **interleaved latent visual reasoning framework** that lets an MLLM alternate between language tokens and continuous latent visual spans during autoregressive decoding. To train this capability, we construct **Future-L1-50K** by selecting examples where future visual hints help prediction and aligning latent states to future-frame embeddings, then further optimizing sampled latent trajectories with **LA-DAPO**, a **latent-aware RL** objective with outcome-contrastive and temporal-diversity rewards. **Future-L1** achieves new **state-of-the-art** results on both benchmarks: on FutureBench, it improves Qwen3-VL-8B from 61.0 to **85.4** and exceeds the previous best Video-CoE by 10.4 points; on TwiFF-Bench, it improves the average score from 2.44 to **3.04**. These results suggest that future-oriented video reasoning benefits from preserving intermediate visual semantics in latent space rather than translating every reasoning step into text.</abstract>
<identifier type="citekey">jiang-etal-2026-imagine</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2315/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>52600</start>
<end>52621</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Imagine Before You Predict: Interleaved Latent Visual Reasoning for Video Event Prediction
%A Jiang, Tianxiang
%A Wu, Linquan
%A Xia, Sheng
%A Li, Songze
%A Yan, Ziang
%A Yang, Haoyu
%A Qiao, Yu
%A Wang, Yi
%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 jiang-etal-2026-imagine
%X Video event prediction (VEP) requires models to infer unobserved future states from partial video evidence. Existing video MLLMs usually verbalize intermediate future reasoning in text space: once visual evidence is verbalized, fine-grained motion, geometry, and interaction cues can be lost, leading to plausible but visually ungrounded hallucinations. We introduce **Future-L1**, an **interleaved latent visual reasoning framework** that lets an MLLM alternate between language tokens and continuous latent visual spans during autoregressive decoding. To train this capability, we construct **Future-L1-50K** by selecting examples where future visual hints help prediction and aligning latent states to future-frame embeddings, then further optimizing sampled latent trajectories with **LA-DAPO**, a **latent-aware RL** objective with outcome-contrastive and temporal-diversity rewards. **Future-L1** achieves new **state-of-the-art** results on both benchmarks: on FutureBench, it improves Qwen3-VL-8B from 61.0 to **85.4** and exceeds the previous best Video-CoE by 10.4 points; on TwiFF-Bench, it improves the average score from 2.44 to **3.04**. These results suggest that future-oriented video reasoning benefits from preserving intermediate visual semantics in latent space rather than translating every reasoning step into text.
%U https://aclanthology.org/2026.emnlp-main.2315/
%P 52600-52621
Markdown (Informal)
[Imagine Before You Predict: Interleaved Latent Visual Reasoning for Video Event Prediction](https://aclanthology.org/2026.emnlp-main.2315/) (Jiang et al., EMNLP 2026)
ACL
- Tianxiang Jiang, Linquan Wu, Sheng Xia, Songze Li, Ziang Yan, Haoyu Yang, Yu Qiao, and Yi Wang. 2026. Imagine Before You Predict: Interleaved Latent Visual Reasoning for Video Event Prediction. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 52600–52621, Budapest, Hungary. Association for Computational Linguistics.