ADEPT: An Entropy-Driven Dual-Strategy Agent for Interactive Video Retrieval

Jingwei Xiong; Jundong Liu; Ke Chen; Liang Xu; Shengyuan Han; Yongfeng Huang; Yujin Zhu

arxiv: 2606.28326 · v1 · pith:RHTBPHGRnew · submitted 2026-05-07 · 💻 cs.IR · cs.AI

ADEPT: An Entropy-Driven Dual-Strategy Agent for Interactive Video Retrieval

Ke Chen , Shengyuan Han , Yongfeng Huang , Yujin Zhu , Jingwei Xiong , Liang Xu , Jundong Liu This is my paper

Pith reviewed 2026-06-30 23:17 UTC · model grok-4.3

classification 💻 cs.IR cs.AI

keywords interactive video retrievalentropy-driven decisionASK and REFINE strategiesintent-query gaptraining-free agentdialogue-based retrievalvideo search

0 comments

The pith

ADEPT uses entropy from the retrieval state to decide whether to ask the user or refine the query internally, closing the intent-query gap in video search.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents ADEPT as a training-free agent that measures entropy in the current retrieval state to choose between two strategies: asking the user for more input or refining the query on its own. This setup targets the intent-query gap, where a single text query often fails to express complex intentions when searching large video collections. The approach replaces fixed single-round retrieval with a dynamic dialogue that switches strategies based on that entropy signal. Experiments on two datasets show the method beats non-interactive baselines, heuristic methods, and Video-LLM systems. Readers would care because it offers a simple, no-training way to make retrieval more accurate when queries start vague.

Core claim

ADEPT pioneers an entropy-driven decision engine to efficiently guide dialogue by dynamically selecting between ASK and REFINE strategies. The engine computes entropy from the retrieval state and uses the value to pick the more effective action without any training or domain tuning. This produces an efficient and interpretable interactive strategy that sets a new performance benchmark on two challenging video retrieval datasets.

What carries the argument

The entropy-driven decision engine that computes an entropy value from the current retrieval state and uses it to select between the ASK strategy (query the user) and the REFINE strategy (adjust the query internally).

If this is right

The dual ASK-REFINE choice improves results over any single fixed strategy on the tested datasets.
The method requires no training data or parameter tuning to reach its reported gains.
The entropy signal supplies an interpretable reason for each strategy switch during a session.
Performance gains hold against both non-interactive baselines and existing Video-LLM approaches.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same entropy switch could be tested in image or text retrieval sessions where initial queries are also ambiguous.
Integrating the engine into existing search interfaces would let systems avoid unnecessary user questions when internal refinement is likely to suffice.
Real-user logs could check whether the entropy-based choices align with what people actually prefer when given the option to answer or not.
If the entropy signal proves stable across datasets, it could serve as a lightweight alternative to learned policy models in other dialogue agents.

Load-bearing premise

An entropy value computed from the retrieval state can reliably indicate whether asking the user or refining internally will produce better results.

What would settle it

A controlled test set where the entropy threshold consistently selects the worse of the two strategies and overall retrieval accuracy drops compared with always using one fixed strategy.

read the original abstract

This research aims to solve the challenge of video retrieval from massive datasets, caused by ambiguous user queries. Prevailing single-round retrieval paradigms face a performance bottleneck, as they lack effective feedback mechanisms to handle complex search intentions. The root cause is the "Intent-Query Gap", where users' intent cannot be captured by a simple text query. To solve this, we propose the ADEPT framework: a training-free agent that pioneers an entropy-driven decision engine to efficiently guide dialogue by dynamically selecting between ASK and REFINE strategies. Experiments on two challenging datasets demonstrate that ADEPT significantly outperforms all non-interactive, heuristic, and Video-LLM baselines. The core contribution of this work is an efficient and interpretable entropy-driven interactive strategy that sets a new performance benchmark for the field of interactive video retrieval.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

ADEPT puts forward a training-free entropy switch to pick between asking the user and refining the query in video retrieval, but the performance claims rest on details not visible in the abstract.

read the letter

The main takeaway is that this paper describes a training-free agent that computes an entropy score from the current retrieval state and uses it to choose between an ASK strategy and a REFINE strategy during interactive video search.

What is new is the framing of the interaction decision as an entropy-driven choice between those two actions. The approach stays training-free, which removes the need for labeled dialogue data or fine-tuning, and it targets the intent-query gap directly.

The paper does a reasonable job stating the problem and reporting that ADEPT beats non-interactive, heuristic, and Video-LLM baselines on two datasets. That gives a concrete point of comparison if the numbers hold.

The soft spots are the missing experimental details. The abstract supplies no entropy formula, no description of how the two strategies are implemented, no dataset names or sizes, and no ablation results. Without those, it is hard to judge whether the entropy signal actually drives the gains or whether the decision rule works reliably across cases. The assumption that a raw entropy measure can pick ASK over REFINE without any tuning remains untested in the visible material.

This paper is aimed at people working on interactive multimedia retrieval who want a lightweight heuristic. A reader looking for a simple, training-free idea might find it worth trying, but the current write-up leaves the central claim difficult to assess.

I would send it for peer review so the authors can supply the missing implementation and results.

Referee Report

1 major / 0 minor

Summary. The paper proposes ADEPT, a training-free agent for interactive video retrieval that uses an entropy-driven decision engine to dynamically select between ASK and REFINE strategies in order to bridge the Intent-Query Gap, claiming significant outperformance over non-interactive, heuristic, and Video-LLM baselines on two challenging datasets.

Significance. A validated entropy-based mechanism that reliably chooses between user interaction and internal refinement without training or tuning could offer an efficient, interpretable advance for handling ambiguous queries in video retrieval. The absence of any experimental details, metrics, datasets, or ablations in the manuscript prevents evaluation of whether this potential is realized.

major comments (1)

[Abstract] Abstract: the central claim that ADEPT 'significantly outperforms all non-interactive, heuristic, and Video-LLM baselines on two challenging datasets' is asserted with no accompanying metrics, dataset names, experimental protocol, results tables, or ablation studies, so the empirical result cannot be checked against the claim.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their review. The single major comment is addressed below.

read point-by-point responses

Referee: [Abstract] Abstract: the central claim that ADEPT 'significantly outperforms all non-interactive, heuristic, and Video-LLM baselines on two challenging datasets' is asserted with no accompanying metrics, dataset names, experimental protocol, results tables, or ablation studies, so the empirical result cannot be checked against the claim.

Authors: We agree that the provided manuscript text contains only the abstract and lacks the supporting experimental details, metrics, dataset names, protocol, tables, or ablations. This prevents verification of the claim. We will revise the manuscript to add the missing experimental section, including dataset names, metrics, protocol description, results tables, and ablations. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The supplied abstract and description contain no equations, derivations, or self-citations that could be inspected for reduction to inputs by construction. The entropy-driven engine is presented as an input mechanism for strategy selection rather than a quantity derived from or fitted to the target performance metric. No load-bearing step reduces to a self-definition, fitted prediction, or author-imported uniqueness theorem. The central claim rests on empirical outperformance on external datasets, which is independent of the method's internal justification.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Review is based solely on the abstract; the ledger is therefore incomplete and provisional.

axioms (1)

domain assumption An Intent-Query Gap exists that interactive dialogue can close more effectively than single-round retrieval.
Explicitly stated in the abstract as the root cause of the performance bottleneck.

invented entities (1)

entropy-driven decision engine no independent evidence
purpose: Dynamically select between ASK and REFINE strategies
Presented as the central technical contribution of the framework.

pith-pipeline@v0.9.1-grok · 5676 in / 1191 out tokens · 21674 ms · 2026-06-30T23:17:19.785133+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

21 extracted references · 3 canonical work pages · 3 internal anchors

[1]

Intent-Query Gap

INTRODUCTION The exponential growth of short-form video creates a signifi- cant “Intent-Query Gap” where users’ vague episodic memo- ries clash with the ambiguity of textual queries. While inter- active retrieval shifts the paradigm from static matching (e.g., CLIP4Clip [1]) to multi-turn dialogue, it exposes a critical “Strategy Gap”: the lack of an opti...
[2]

ADEPT: An Entropy-Driven Dual-Strategy Agent for Interactive Video Retrieval

RELATED WORKS Interactive video retrieval bridges the “Intent-Query Gap” through multi-turn dialogue. This shifts the research paradigm from static semantic matching to defining an optimal dialogue policy for the “Strategy Gap”. Current approaches are mainly categorized into non-interactive retrieval, heuristic-based in- teraction, and agent-based systems...

work page internal anchor Pith review Pith/arXiv arXiv 2026
[3]

METHODOLOGY We propose ADEPT, a training-free framework that interprets user intent through entropy-driven interactions. 3.1. Overview of the Framework Given a databaseDand queryD 0, the system iteratively re- fines results overTrounds. We map inputs to a shared seman- tic space via an encoderf enc(·). At roundt, the queryD t re- trieves the candidate set...
[4]

EXPERIMENTAL RESULTS This section presents a comprehensive empirical evaluation of the proposed ADEPT framework. Fig. 1. Overview of the ADEPT framework. The agent operates as an entropy-drivenclosed loop, utilizing uncertainty metrics (Hinter, Hintra) to dynamically select optimal strategies. Specificsystem promptsfor each VLM agent are detailed in the c...
[5]

CONCLUSION We present ADEPT, a training-free, entropy-driven agent de- signed to bridge the intent—query gap. By diagnosing un- certainty via inter- and intra-cluster entropy, ADEPT dynami- cally switches strategies to efficiently guide interaction, estab- lishing a principled baseline for interpretable active retrieval. Limitations & Future Work.ADEPT cu...
[6]

CLIP4CLIP: An em- pirical study of clip for end to end video clip retrieval,

Huaishao Luo, Lei Ji, Ming Zhong, Yang Chen, Wen Lei, Nan Duan, and Tianrui Li, “CLIP4CLIP: An em- pirical study of clip for end to end video clip retrieval,” Neurocomputing, vol. 508, pp. 293–304, 2021

2021
[7]

TraveLER: A modular multi- LMM agent framework for video question-answering,

Chuyi Shang, Amos You, Sanjay Subramanian, Trevor Darrell, and Roei Herzig, “TraveLER: A modular multi- LMM agent framework for video question-answering,” inProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, Yaser Al- Onaizan, Mohit Bansal, and Yun-Nung Chen, Eds., Mi- ami, Florida, USA, Nov. 2024, pp. 9740–9766, Assoc...

2024
[8]

MERLIN: Multimodal embedding refinement via LLM-based iterative navigation for text- video retrieval-rerank pipeline,

Donghoon Han, Eunhwan Park, Gisang Lee, Adam Lee, and Nojun Kwak, “MERLIN: Multimodal embedding refinement via LLM-based iterative navigation for text- video retrieval-rerank pipeline,” inProceedings of the 2024 Conference on Empirical Methods in Natural Lan- guage Processing: Industry Track, Franck Dernoncourt, Daniel Preot ¸iuc-Pietro, and Anastasia Shi...

2024
[9]

Videoagent: A memory- augmented multimodal agent for video understanding,

Yue Fan, Xiaojian Ma, Rujie Wu, Yuntao Du, Jiaqi Li, Zhi Gao, and Qing Li, “Videoagent: A memory- augmented multimodal agent for video understanding,” inProceedings of the European Conference on Com- puter Vision, 2024, pp. 75–92

2024
[10]

Inter- active video retrieval with dialog,

Sho Maeoki, Kohei Uehara, and Tatsuya Harada, “Inter- active video retrieval with dialog,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2020, pp. 952–953

2020
[11]

Simple baselines for interactive video retrieval with questions and answers,

Kaiqu Liang and Samuel Albanie, “Simple baselines for interactive video retrieval with questions and answers,” inProceedings of the IEEE/CVF International Confer- ence on Computer Vision, 2023, pp. 11091–11101

2023
[12]

MAFW: A large-scale, multi-modal, compound affec- tive database for dynamic facial expression recognition in the wild,

Yuanyuan Liu, Wei Dai, Chuanxu Feng, Wenbin Wang, Guanghao Yin, Jiabei Zeng, and Shiguang Shan, “MAFW: A large-scale, multi-modal, compound affec- tive database for dynamic facial expression recognition in the wild,” inProceedings of the 30th ACM Interna- tional Conference on Multimedia, 2022, pp. 24–32

2022
[13]

MER 2024: Semi- supervised learning, noise robustness, and open- vocabulary multimodal emotion recognition,

Zheng Lian, Haiyang Sun, Licai Sun, Zhuofan Wen, Siyuan Zhang, Shun Chen, Hao Gu, Jinming Zhao, Ziyang Ma, Xie Chen, et al., “MER 2024: Semi- supervised learning, noise robustness, and open- vocabulary multimodal emotion recognition,” inPro- ceedings of the 2nd International Workshop on Multi- modal and Responsible Affective Computing, 2024, pp. 41–48

2024
[14]

X-CLIP: End-to-end multi- grained contrastive learning for video-text retrieval,

Yiwei Ma, Guohai Xu, Xiaoshuai Sun, Ming Yan, Ji Zhang, and Rongrong Ji, “X-CLIP: End-to-end multi- grained contrastive learning for video-text retrieval,” in Proceedings of the 30th ACM International Conference on Multimedia, 2022, pp. 638–647

2022
[15]

Emotion-LLaMA: Multimodal emo- tion recognition and reasoning with instruction tuning,

Zebang Cheng, Zhi-Qi Cheng, Jun-Yan He, Kai Wang, Yuxiang Lin, Zheng Lian, Xiaojiang Peng, and Alexan- der Hauptmann, “Emotion-LLaMA: Multimodal emo- tion recognition and reasoning with instruction tuning,” Advances in Neural Information Processing Systems, vol. 37, pp. 110805–110853, 2024

2024
[16]

Composed video re- trieval via enriched context and discriminative embed- dings,

Omkar Thawakar, Muzammal Naseer, Rao Muham- mad Anwer, Salman Khan, Michael Felsberg, Mubarak Shah, and Fahad Shahbaz Khan, “Composed video re- trieval via enriched context and discriminative embed- dings,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 26896–26906

2024
[17]

LLaVA-NeXT-Interleave: Tackling Multi-image, Video, and 3D in Large Multimodal Models

Feng Li, Renrui Zhang, Hao Zhang, Yuanhan Zhang, Bo Li, Wei Li, Zejun Ma, and Chunyuan Li, “LLaV A- NeXT-Interleave: Tackling multi-image, video, and 3d in large multimodal models,”arXiv preprint arXiv:2407.07895, 2024

work page internal anchor Pith review Pith/arXiv arXiv 2024
[18]

A mathematical theory of commu- nication,

Claude E Shannon, “A mathematical theory of commu- nication,”The Bell system technical journal, vol. 27, no. 3, pp. 379–423, 1948

1948
[19]

Qwen2.5-VL Technical Report

Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wen- bin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, et al., “Qwen2.5-VL technical report,”arXiv preprint arXiv:2502.13923, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025
[20]

Msr-vtt: A large video description dataset for bridging video and language,

Jun Xu, Tao Mei, Ting Yao, and Yong Rui, “Msr-vtt: A large video description dataset for bridging video and language,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 5288–5296

2016
[21]

MBA-RAG: a bandit approach for adaptive retrieval-augmented generation through ques- tion complexity,

Xiaqiang Tang, Qiang Gao, Jian Li, Nan Du, Qi Li, and Sihong Xie, “MBA-RAG: a bandit approach for adaptive retrieval-augmented generation through ques- tion complexity,” inProceedings of the 31st Interna- tional Conference on Computational Linguistics, 2025, pp. 3248–3254. Compliance with Ethical Standards This research did not involve human participants ...

2025

[1] [1]

Intent-Query Gap

INTRODUCTION The exponential growth of short-form video creates a signifi- cant “Intent-Query Gap” where users’ vague episodic memo- ries clash with the ambiguity of textual queries. While inter- active retrieval shifts the paradigm from static matching (e.g., CLIP4Clip [1]) to multi-turn dialogue, it exposes a critical “Strategy Gap”: the lack of an opti...

[2] [2]

ADEPT: An Entropy-Driven Dual-Strategy Agent for Interactive Video Retrieval

RELATED WORKS Interactive video retrieval bridges the “Intent-Query Gap” through multi-turn dialogue. This shifts the research paradigm from static semantic matching to defining an optimal dialogue policy for the “Strategy Gap”. Current approaches are mainly categorized into non-interactive retrieval, heuristic-based in- teraction, and agent-based systems...

work page internal anchor Pith review Pith/arXiv arXiv 2026

[3] [3]

METHODOLOGY We propose ADEPT, a training-free framework that interprets user intent through entropy-driven interactions. 3.1. Overview of the Framework Given a databaseDand queryD 0, the system iteratively re- fines results overTrounds. We map inputs to a shared seman- tic space via an encoderf enc(·). At roundt, the queryD t re- trieves the candidate set...

[4] [4]

EXPERIMENTAL RESULTS This section presents a comprehensive empirical evaluation of the proposed ADEPT framework. Fig. 1. Overview of the ADEPT framework. The agent operates as an entropy-drivenclosed loop, utilizing uncertainty metrics (Hinter, Hintra) to dynamically select optimal strategies. Specificsystem promptsfor each VLM agent are detailed in the c...

[5] [5]

CONCLUSION We present ADEPT, a training-free, entropy-driven agent de- signed to bridge the intent—query gap. By diagnosing un- certainty via inter- and intra-cluster entropy, ADEPT dynami- cally switches strategies to efficiently guide interaction, estab- lishing a principled baseline for interpretable active retrieval. Limitations & Future Work.ADEPT cu...

[6] [6]

CLIP4CLIP: An em- pirical study of clip for end to end video clip retrieval,

Huaishao Luo, Lei Ji, Ming Zhong, Yang Chen, Wen Lei, Nan Duan, and Tianrui Li, “CLIP4CLIP: An em- pirical study of clip for end to end video clip retrieval,” Neurocomputing, vol. 508, pp. 293–304, 2021

2021

[7] [7]

TraveLER: A modular multi- LMM agent framework for video question-answering,

Chuyi Shang, Amos You, Sanjay Subramanian, Trevor Darrell, and Roei Herzig, “TraveLER: A modular multi- LMM agent framework for video question-answering,” inProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, Yaser Al- Onaizan, Mohit Bansal, and Yun-Nung Chen, Eds., Mi- ami, Florida, USA, Nov. 2024, pp. 9740–9766, Assoc...

2024

[8] [8]

MERLIN: Multimodal embedding refinement via LLM-based iterative navigation for text- video retrieval-rerank pipeline,

Donghoon Han, Eunhwan Park, Gisang Lee, Adam Lee, and Nojun Kwak, “MERLIN: Multimodal embedding refinement via LLM-based iterative navigation for text- video retrieval-rerank pipeline,” inProceedings of the 2024 Conference on Empirical Methods in Natural Lan- guage Processing: Industry Track, Franck Dernoncourt, Daniel Preot ¸iuc-Pietro, and Anastasia Shi...

2024

[9] [9]

Videoagent: A memory- augmented multimodal agent for video understanding,

Yue Fan, Xiaojian Ma, Rujie Wu, Yuntao Du, Jiaqi Li, Zhi Gao, and Qing Li, “Videoagent: A memory- augmented multimodal agent for video understanding,” inProceedings of the European Conference on Com- puter Vision, 2024, pp. 75–92

2024

[10] [10]

Inter- active video retrieval with dialog,

Sho Maeoki, Kohei Uehara, and Tatsuya Harada, “Inter- active video retrieval with dialog,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2020, pp. 952–953

2020

[11] [11]

Simple baselines for interactive video retrieval with questions and answers,

Kaiqu Liang and Samuel Albanie, “Simple baselines for interactive video retrieval with questions and answers,” inProceedings of the IEEE/CVF International Confer- ence on Computer Vision, 2023, pp. 11091–11101

2023

[12] [12]

MAFW: A large-scale, multi-modal, compound affec- tive database for dynamic facial expression recognition in the wild,

Yuanyuan Liu, Wei Dai, Chuanxu Feng, Wenbin Wang, Guanghao Yin, Jiabei Zeng, and Shiguang Shan, “MAFW: A large-scale, multi-modal, compound affec- tive database for dynamic facial expression recognition in the wild,” inProceedings of the 30th ACM Interna- tional Conference on Multimedia, 2022, pp. 24–32

2022

[13] [13]

MER 2024: Semi- supervised learning, noise robustness, and open- vocabulary multimodal emotion recognition,

Zheng Lian, Haiyang Sun, Licai Sun, Zhuofan Wen, Siyuan Zhang, Shun Chen, Hao Gu, Jinming Zhao, Ziyang Ma, Xie Chen, et al., “MER 2024: Semi- supervised learning, noise robustness, and open- vocabulary multimodal emotion recognition,” inPro- ceedings of the 2nd International Workshop on Multi- modal and Responsible Affective Computing, 2024, pp. 41–48

2024

[14] [14]

X-CLIP: End-to-end multi- grained contrastive learning for video-text retrieval,

Yiwei Ma, Guohai Xu, Xiaoshuai Sun, Ming Yan, Ji Zhang, and Rongrong Ji, “X-CLIP: End-to-end multi- grained contrastive learning for video-text retrieval,” in Proceedings of the 30th ACM International Conference on Multimedia, 2022, pp. 638–647

2022

[15] [15]

Emotion-LLaMA: Multimodal emo- tion recognition and reasoning with instruction tuning,

Zebang Cheng, Zhi-Qi Cheng, Jun-Yan He, Kai Wang, Yuxiang Lin, Zheng Lian, Xiaojiang Peng, and Alexan- der Hauptmann, “Emotion-LLaMA: Multimodal emo- tion recognition and reasoning with instruction tuning,” Advances in Neural Information Processing Systems, vol. 37, pp. 110805–110853, 2024

2024

[16] [16]

Composed video re- trieval via enriched context and discriminative embed- dings,

Omkar Thawakar, Muzammal Naseer, Rao Muham- mad Anwer, Salman Khan, Michael Felsberg, Mubarak Shah, and Fahad Shahbaz Khan, “Composed video re- trieval via enriched context and discriminative embed- dings,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 26896–26906

2024

[17] [17]

LLaVA-NeXT-Interleave: Tackling Multi-image, Video, and 3D in Large Multimodal Models

Feng Li, Renrui Zhang, Hao Zhang, Yuanhan Zhang, Bo Li, Wei Li, Zejun Ma, and Chunyuan Li, “LLaV A- NeXT-Interleave: Tackling multi-image, video, and 3d in large multimodal models,”arXiv preprint arXiv:2407.07895, 2024

work page internal anchor Pith review Pith/arXiv arXiv 2024

[18] [18]

A mathematical theory of commu- nication,

Claude E Shannon, “A mathematical theory of commu- nication,”The Bell system technical journal, vol. 27, no. 3, pp. 379–423, 1948

1948

[19] [19]

Qwen2.5-VL Technical Report

Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wen- bin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, et al., “Qwen2.5-VL technical report,”arXiv preprint arXiv:2502.13923, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025

[20] [20]

Msr-vtt: A large video description dataset for bridging video and language,

Jun Xu, Tao Mei, Ting Yao, and Yong Rui, “Msr-vtt: A large video description dataset for bridging video and language,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 5288–5296

2016

[21] [21]

MBA-RAG: a bandit approach for adaptive retrieval-augmented generation through ques- tion complexity,

Xiaqiang Tang, Qiang Gao, Jian Li, Nan Du, Qi Li, and Sihong Xie, “MBA-RAG: a bandit approach for adaptive retrieval-augmented generation through ques- tion complexity,” inProceedings of the 31st Interna- tional Conference on Computational Linguistics, 2025, pp. 3248–3254. Compliance with Ethical Standards This research did not involve human participants ...

2025