pith. sign in

arxiv: 2606.11074 · v2 · pith:GWTT5ZHVnew · submitted 2026-06-09 · 💻 cs.CL · cs.AI

Modeling Complex Behaviors: Multi-Personality Composition and Dynamic Switching in Vision-Language Models

Peiqi Jia , Haonan Jia , Ziqi Miao , Linkang Du , Yuntao Wang , Zhou Su , (2) Beihang University) This is my paper

Pith reviewed 2026-06-27 13:02 UTC · model grok-4.3

classification 💻 cs.CL cs.AI
keywords personality inductionmultimodal large language modelsvision-language modelsmulti-personality compositiondynamic switchingimage captioningvisual question answeringbehavior modulation
0
0 comments X

The pith

Prompt-based personality induction in vision-language models improves image captioning but impairs performance on reasoning tasks like VQA.

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

The paper introduces explicit personality conditioning for multimodal large language models and builds an evaluation framework covering single-personality induction, multi-personality composition, and dynamic switching. Experiments demonstrate that these prompts enhance descriptive output on captioning tasks while degrading accuracy on visual question answering that demands precise reasoning. Balancing effects appear when traits combine and residual effects persist when traits switch, showing that prior constraints continue to shape responses alongside new ones. This matters for deployed models in social settings where behavior must be predictable and controllable. The findings also indicate that text-only prompt methods transfer poorly to multimodal cases.

Core claim

Model behavior is co-modulated by both previous and current personality constraints, as evidenced by balancing and residual effects observed during multi-trait composition and dynamic switching; personality induction improves image captioning performance but can impair performance on tasks requiring precise reasoning such as visual question answering.

What carries the argument

Explicit personality conditioning via prompts, which induces, composes, and switches personality traits to produce measurable changes in model output on captioning and VQA tasks.

If this is right

  • Personality induction selectively boosts descriptive performance on image captioning.
  • The same induction impairs accuracy on visual question answering that requires precise reasoning.
  • Multi-trait composition produces balancing effects between competing personality constraints.
  • Dynamic switching leaves residual influence from the prior personality on subsequent outputs.
  • Existing prompt-based induction methods show limited transferability from text-only to multimodal settings.

Where Pith is reading between the lines

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

  • Stable personality control in interactive applications may require methods beyond simple prompting.
  • Residual effects could complicate long-running conversations where users expect consistent traits.
  • The observed task-specific trade-offs suggest designers must choose induction targets based on primary use case.
  • Limited multimodal transfer points to the value of testing induction techniques directly on image-text data rather than assuming text results generalize.

Load-bearing premise

Prompt-based personality induction produces measurable, stable personality-like behavior in MLLMs that can be isolated from other prompt or model effects without confounding factors in the chosen tasks.

What would settle it

Running the same captioning and VQA evaluations with personality prompts versus matched neutral prompts and finding no consistent performance differences or balancing/residual patterns would falsify the claim of distinct co-modulated effects.

Figures

Figures reproduced from arXiv: 2606.11074 by (2) Beihang University), Haonan Jia, Linkang Du, Peiqi Jia, Yuntao Wang, Zhou Su, Ziqi Miao.

Figure 1
Figure 1. Figure 1: Illustration of personality induction. The top-left panel illustrates generation under single-personality [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of single-personality, multi-personality induction, and in-conversation personality switching. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Personality assessment results of the Qwen2.5-VL-7B model based on the BFI and IPIP-NEO. We report [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Personality assessment under in-conversation personality switching settings using BFI and IPIP-NEO. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
read the original abstract

With the widespread deployment of Multimodal Large Language Models (MLLMs) in social interaction, understanding and controlling their behavior under complex personality conditions is essential. This paper introduces explicit personality conditioning and establishes a systematic evaluation framework encompassing single-personality induction, multi-personality induction, and personality switching. Experiments show that personality induction improves image captioning performance but can impair performance on tasks requiring precise reasoning, such as visual question answering (VQA). Balancing and residual effects are observed during multi-trait composition and dynamic switching, indicating that model behavior is co-modulated by both previous and current personality constraints. Existing prompt-based personality induction methods show limited transferability to multimodal settings. Our work reveals the dynamic and complex nature of personality modeling in MLLMs and underscores the need for robust, tailored methods for personality induction and evaluation. The code will be released when the paper is accepted.

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.

Referee Report

2 major / 1 minor

Summary. The paper introduces explicit personality conditioning for Multimodal Large Language Models (MLLMs) along with a systematic evaluation framework covering single-personality induction, multi-personality composition, and dynamic personality switching. Experiments are reported to show that personality induction improves image captioning performance while impairing precise reasoning tasks such as VQA; balancing and residual effects are observed in multi-trait and switching scenarios, indicating co-modulation by prior and current personality constraints. Existing prompt-based methods exhibit limited transferability to multimodal settings, and the work calls for more robust induction and evaluation techniques. Code release is promised upon acceptance.

Significance. If the directional effects and balancing/residual phenomena are shown to be robust under controlled conditions with appropriate statistical support, the work would usefully illustrate the complexity of behavioral modulation in MLLMs and the shortcomings of current prompt-based approaches. This could help guide research on controllable multimodal systems for social applications. The commitment to release code supports reproducibility.

major comments (2)
  1. [Abstract] Abstract: The central claims of performance improvement on captioning, impairment on VQA, and the existence of balancing/residual effects are presented as directional observations without any reported sample sizes, number of trials, statistical tests, baselines, or error bars. This absence is load-bearing because the evaluation framework and all subsequent claims rest on these unspecified experiments.
  2. [Evaluation framework] Evaluation framework and experimental sections: The premise that prompt-based personality induction produces isolatable, stable personality-like behavior (distinct from general prompt or model effects) is not supported by described controls, ablations, or validation metrics. This directly affects interpretation of the multi-personality composition and switching results.
minor comments (1)
  1. The statement that code will be released upon acceptance is welcome; consider adding a footnote or repository link in the final version for immediate access.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback. We address each major comment below and indicate planned revisions to the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claims of performance improvement on captioning, impairment on VQA, and the existence of balancing/residual effects are presented as directional observations without any reported sample sizes, number of trials, statistical tests, baselines, or error bars. This absence is load-bearing because the evaluation framework and all subsequent claims rest on these unspecified experiments.

    Authors: We agree that the abstract and supporting experimental descriptions would benefit from explicit statistical reporting. In the revised manuscript we will update the abstract to reference sample sizes, number of trials, baselines, error bars, and the statistical tests performed in the main text, and we will expand the evaluation framework section to include these details for all reported effects. revision: yes

  2. Referee: [Evaluation framework] Evaluation framework and experimental sections: The premise that prompt-based personality induction produces isolatable, stable personality-like behavior (distinct from general prompt or model effects) is not supported by described controls, ablations, or validation metrics. This directly affects interpretation of the multi-personality composition and switching results.

    Authors: We acknowledge that the current description relies primarily on downstream task differences to infer personality effects. We will add explicit controls and ablations (including neutral-prompt baselines and consistency checks across prompt variants) plus validation metrics in the revised experimental sections to better isolate personality-specific modulation from general prompt or model effects. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper is strictly empirical: it describes an evaluation framework for personality induction, multi-personality composition, and switching in MLLMs, then reports observed performance shifts on captioning and VQA tasks along with balancing/residual effects. No derivation chain, equations, fitted parameters, or first-principles predictions appear; all central claims are direct statements of experimental outcomes rather than reductions to inputs or self-citations. The work is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central observations rest on the unstated assumption that personality can be reliably induced and measured via prompts in MLLMs without additional model-specific tuning or task artifacts. No free parameters, invented entities, or explicit axioms are stated in the abstract.

axioms (1)
  • domain assumption Prompt-based conditioning can isolate and control personality-like behavior in MLLMs independently of other prompt effects.
    Invoked implicitly when the abstract treats observed performance changes as evidence of personality induction and composition effects.

pith-pipeline@v0.9.1-grok · 5699 in / 1256 out tokens · 15312 ms · 2026-06-27T13:02:24.220054+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

105 extracted references · 2 canonical work pages

  1. [1]

    2016 , eprint=

    VQA: Visual Question Answering , author=. 2016 , eprint=

    2016

  2. [2]

    2019 , eprint=

    ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks , author=. 2019 , eprint=

    2019

  3. [3]

    2023 , eprint=

    Building Emotional Support Chatbots in the Era of LLMs , author=. 2023 , eprint=

    2023

  4. [4]

    SSRN Electronic Journal , year =

    ChatGPT User Experience: Implications for Education , author =. SSRN Electronic Journal , year =

  5. [5]

    2020 , eprint=

    DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation , author=. 2020 , eprint=

    2020

  6. [6]

    2023 , eprint=

    Sparks of Artificial General Intelligence: Early experiments with GPT-4 , author=. 2023 , eprint=

    2023

  7. [8]

    2015 , eprint=

    Exploring Models and Data for Image Question Answering , author=. 2015 , eprint=

    2015

  8. [9]

    2016 , eprint=

    Stacked Attention Networks for Image Question Answering , author=. 2016 , eprint=

    2016

  9. [10]

    2018 , eprint=

    Bottom-Up and Top-Down Attention for Image Captioning and Visual Question Answering , author=. 2018 , eprint=

    2018

  10. [11]

    2021 , eprint=

    Learning Transferable Visual Models From Natural Language Supervision , author=. 2021 , eprint=

    2021

  11. [12]

    2021 , eprint=

    Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision , author=. 2021 , eprint=

    2021

  12. [13]

    2021 , eprint=

    Masked Autoencoders Are Scalable Vision Learners , author=. 2021 , eprint=

    2021

  13. [14]

    2022 , eprint=

    BEiT v2: Masked Image Modeling with Vector-Quantized Visual Tokenizers , author=. 2022 , eprint=

    2022

  14. [15]

    2022 , eprint=

    BEiT: BERT Pre-Training of Image Transformers , author=. 2022 , eprint=

    2022

  15. [16]

    2023 , eprint=

    Visual Instruction Tuning , author=. 2023 , eprint=

    2023

  16. [17]

    2023 , eprint=

    Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond , author=. 2023 , eprint=

    2023

  17. [18]

    2024 , eprint=

    Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution , author=. 2024 , eprint=

    2024

  18. [19]

    2024 , eprint=

    DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understanding , author=. 2024 , eprint=

    2024

  19. [20]

    2025 , eprint=

    Qwen3 Technical Report , author=. 2025 , eprint=

    2025

  20. [21]

    2021 , eprint=

    ClipCap: CLIP Prefix for Image Captioning , author=. 2021 , eprint=

    2021

  21. [22]

    2022 , eprint=

    ZeroCap: Zero-Shot Image-to-Text Generation for Visual-Semantic Arithmetic , author=. 2022 , eprint=

    2022

  22. [23]

    2025 , eprint=

    SC-Captioner: Improving Image Captioning with Self-Correction by Reinforcement Learning , author=. 2025 , eprint=

    2025

  23. [24]

    2024 , eprint=

    MMBench: Is Your Multi-modal Model an All-around Player? , author=. 2024 , eprint=

    2024

  24. [25]

    2025 , eprint=

    MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models , author=. 2025 , eprint=

    2025

  25. [26]

    2025 , eprint=

    Qwen2.5-Omni Technical Report , author=. 2025 , eprint=

    2025

  26. [27]

    The Big Five Inventory---Versions 4a and 54 , author =

  27. [28]

    1998 , publisher =

    MBTI Manual: A Guide to the Development and Use of the Myers-Briggs Type Indicator , author =. 1998 , publisher =

    1998

  28. [29]

    2024 , eprint=

    Self-assessment, Exhibition, and Recognition: a Review of Personality in Large Language Models , author=. 2024 , eprint=

    2024

  29. [30]

    2025 , eprint=

    Persona-Assigned Large Language Models Exhibit Human-Like Motivated Reasoning , author=. 2025 , eprint=

    2025

  30. [31]

    2024 , eprint=

    HallusionBench: An Advanced Diagnostic Suite for Entangled Language Hallucination and Visual Illusion in Large Vision-Language Models , author=. 2024 , eprint=

    2024

  31. [32]

    2025 , eprint=

    Qwen2.5-VL Technical Report , author=. 2025 , eprint=

    2025

  32. [33]

    2024 , eprint=

    MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts , author=. 2024 , eprint=

    2024

  33. [34]

    2024 , eprint=

    Are We on the Right Way for Evaluating Large Vision-Language Models? , author=. 2024 , eprint=

    2024

  34. [35]

    2023 , eprint=

    SEED-Bench: Benchmarking Multimodal LLMs with Generative Comprehension , author=. 2023 , eprint=

    2023

  35. [36]

    2024 , eprint=

    MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI , author=. 2024 , eprint=

    2024

  36. [37]

    On the Humanity of Conversational

    Jen-tse Huang and Wenxuan Wang and Eric John Li and Man Ho LAM and Shujie Ren and Youliang Yuan and Wenxiang Jiao and Zhaopeng Tu and Michael Lyu , booktitle=. On the Humanity of Conversational. 2024 , url=

    2024

  37. [38]

    2025 , eprint=

    BIG5-CHAT: Shaping LLM Personalities Through Training on Human-Grounded Data , author=. 2025 , eprint=

    2025

  38. [39]

    2024 , eprint=

    The Better Angels of Machine Personality: How Personality Relates to LLM Safety , author=. 2024 , eprint=

    2024

  39. [40]

    2025 , eprint=

    Psychometric Personality Shaping Modulates Capabilities and Safety in Language Models , author=. 2025 , eprint=

    2025

  40. [41]

    2025 , eprint=

    Personality Alignment of Large Language Models , author=. 2025 , eprint=

    2025

  41. [42]

    2023 , eprint=

    Evaluating and Inducing Personality in Pre-trained Language Models , author=. 2023 , eprint=

    2023

  42. [43]

    2025 , eprint=

    Scaling Personality Control in LLMs with Big Five Scaler Prompts , author=. 2025 , eprint=

    2025

  43. [44]

    2024 , eprint=

    Editing Personality for Large Language Models , author=. 2024 , eprint=

    2024

  44. [45]

    2024 , eprint=

    Neuron-based Personality Trait Induction in Large Language Models , author=. 2024 , eprint=

    2024

  45. [46]

    2025 , eprint=

    Personality Vector: Modulating Personality of Large Language Models by Model Merging , author=. 2025 , eprint=

    2025

  46. [47]

    Aho and Jeffrey D

    Alfred V. Aho and Jeffrey D. Ullman , title =. 1972

    1972

  47. [48]

    Publications Manual , year = "1983", publisher =

    1983

  48. [49]

    Chandra and Dexter C

    Ashok K. Chandra and Dexter C. Kozen and Larry J. Stockmeyer , year = "1981", title =. doi:10.1145/322234.322243

    work page doi:10.1145/322234.322243 1981

  49. [50]

    Scalable training of

    Andrew, Galen and Gao, Jianfeng , booktitle=. Scalable training of

  50. [51]

    Dan Gusfield , title =. 1997

    1997

  51. [52]

    Tetreault , title =

    Mohammad Sadegh Rasooli and Joel R. Tetreault , title =. Computing Research Repository , volume =. 2015 , url =

    2015

  52. [53]

    A Framework for Learning Predictive Structures from Multiple Tasks and Unlabeled Data , Volume =

    Ando, Rie Kubota and Zhang, Tong , Issn =. A Framework for Learning Predictive Structures from Multiple Tasks and Unlabeled Data , Volume =. Journal of Machine Learning Research , Month = dec, Numpages =

  53. [54]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Cross-modal Identity Mapping: Minimizing Information Loss in Modality Conversion via Reinforcement Learning , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  54. [57]

    arXiv preprint arXiv:2602.22556 , year=

    Stable adaptive thinking via advantage shaping and length-aware gradient regulation , author=. arXiv preprint arXiv:2602.22556 , year=

    arXiv

  55. [58]

    Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , pages=

    Visual contextual attack: Jailbreaking mllms with image-driven context injection , author=. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , pages=

    2025

  56. [59]

    Lawrence Zitnick, Dhruv Batra, and Devi Parikh

    Aishwarya Agrawal, Jiasen Lu, Stanislaw Antol, Margaret Mitchell, C. Lawrence Zitnick, Dhruv Batra, and Devi Parikh. 2016. https://arxiv.org/abs/1505.00468 Vqa: Visual question answering . Preprint, arXiv:1505.00468

    Pith/arXiv arXiv 2016

  57. [60]

    Peter Anderson, Xiaodong He, Chris Buehler, Damien Teney, Mark Johnson, Stephen Gould, and Lei Zhang. 2018. https://arxiv.org/abs/1707.07998 Bottom-up and top-down attention for image captioning and visual question answering . Preprint, arXiv:1707.07998

    Pith/arXiv arXiv 2018

  58. [61]

    Jinze Bai, Shuai Bai, Shusheng Yang, Shijie Wang, Sinan Tan, Peng Wang, Junyang Lin, Chang Zhou, and Jingren Zhou. 2023. https://arxiv.org/abs/2308.12966 Qwen-vl: A versatile vision-language model for understanding, localization, text reading, and beyond . Preprint, arXiv:2308.12966

    Pith/arXiv arXiv 2023

  59. [62]

    Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, Humen Zhong, Yuanzhi Zhu, Mingkun Yang, Zhaohai Li, Jianqiang Wan, Pengfei Wang, Wei Ding, Zheren Fu, Yiheng Xu, and 8 others. 2025. https://arxiv.org/abs/2502.13923 Qwen2.5-vl technical report . Preprint, arXiv:2502.13923

    Pith/arXiv arXiv 2025

  60. [63]

    Hangbo Bao, Li Dong, Songhao Piao, and Furu Wei. 2022. https://arxiv.org/abs/2106.08254 Beit: Bert pre-training of image transformers . Preprint, arXiv:2106.08254

    Pith/arXiv arXiv 2022

  61. [64]

    Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lundberg, Harsha Nori, Hamid Palangi, Marco Tulio Ribeiro, and Yi Zhang. 2023. https://arxiv.org/abs/2303.12712 Sparks of artificial general intelligence: Early experiments with gpt-4 . Preprint, arXiv:2303.12712

    Pith/arXiv arXiv 2023

  62. [65]

    Lin Chen, Jinsong Li, Xiaoyi Dong, Pan Zhang, Yuhang Zang, Zehui Chen, Haodong Duan, Jiaqi Wang, Yu Qiao, Dahua Lin, and Feng Zhao. 2024. https://arxiv.org/abs/2403.20330 Are we on the right way for evaluating large vision-language models? Preprint, arXiv:2403.20330

    Pith/arXiv arXiv 2024

  63. [66]

    Nuo Chen, Yan Wang, Haiyun Jiang, Deng Cai, Yuhan Li, Ziyang Chen, Longyue Wang, and Jia Li. 2023. https://doi.org/10.18653/v1/2023.findings-emnlp.570 Large language models meet harry potter: A dataset for aligning dialogue agents with characters . In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 8506--8520, Singapore. Assoc...

    work page doi:10.18653/v1/2023.findings-emnlp.570 2023

  64. [67]

    Gunhee Cho and Yun-Gyung Cheong. 2025. https://arxiv.org/abs/2508.06149 Scaling personality control in llms with big five scaler prompts . Preprint, arXiv:2508.06149

    arXiv 2025

  65. [68]

    Spiro, and Aylin Caliskan

    Saloni Dash, Amélie Reymond, Emma S. Spiro, and Aylin Caliskan. 2025. https://arxiv.org/abs/2506.20020 Persona-assigned large language models exhibit human-like motivated reasoning . Preprint, arXiv:2506.20020

    Pith/arXiv arXiv 2025

  66. [69]

    Jia Deng, Tianyi Tang, Yanbin Yin, Wenhao Yang, Wayne Xin Zhao, and Ji-Rong Wen. 2024. https://arxiv.org/abs/2410.12327 Neuron-based personality trait induction in large language models . Preprint, arXiv:2410.12327

    Pith/arXiv arXiv 2024

  67. [70]

    Stephen Fitz, Peter Romero, Steven Basart, Sipeng Chen, and Jose Hernandez-Orallo. 2025. https://arxiv.org/abs/2509.16332 Psychometric personality shaping modulates capabilities and safety in language models . Preprint, arXiv:2509.16332

    arXiv 2025

  68. [71]

    Chaoyou Fu, Peixian Chen, Yunhang Shen, Yulei Qin, Mengdan Zhang, Xu Lin, Jinrui Yang, Xiawu Zheng, Ke Li, Xing Sun, Yunsheng Wu, Rongrong Ji, Caifeng Shan, and Ran He. 2025. https://arxiv.org/abs/2306.13394 Mme: A comprehensive evaluation benchmark for multimodal large language models . Preprint, arXiv:2306.13394

    Pith/arXiv arXiv 2025

  69. [72]

    Tianrui Guan, Fuxiao Liu, Xiyang Wu, Ruiqi Xian, Zongxia Li, Xiaoyu Liu, Xijun Wang, Lichang Chen, Furong Huang, Yaser Yacoob, Dinesh Manocha, and Tianyi Zhou. 2024. https://arxiv.org/abs/2310.14566 Hallusionbench: An advanced diagnostic suite for entangled language hallucination and visual illusion in large vision-language models . Preprint, arXiv:2310.14566

    Pith/arXiv arXiv 2024

  70. [73]

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick. 2021. https://arxiv.org/abs/2111.06377 Masked autoencoders are scalable vision learners . Preprint, arXiv:2111.06377

    Pith/arXiv arXiv 2021

  71. [74]

    Le, Yunhsuan Sung, Zhen Li, and Tom Duerig

    Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V. Le, Yunhsuan Sung, Zhen Li, and Tom Duerig. 2021. https://arxiv.org/abs/2102.05918 Scaling up visual and vision-language representation learning with noisy text supervision . Preprint, arXiv:2102.05918

    arXiv 2021

  72. [75]

    Haonan Jia, Shichao Dong, Xin Dong, Zenghui Sun, Jin Wang, Jinsong Lan, Xiaoyong Zhu, Bo Zheng, and Kaifu Zhang. 2026. Cross-modal identity mapping: Minimizing information loss in modality conversion via reinforcement learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 766--777

    2026

  73. [76]

    Guangyuan Jiang, Manjie Xu, Song-Chun Zhu, Wenjuan Han, Chi Zhang, and Yixin Zhu. 2023. https://arxiv.org/abs/2206.07550 Evaluating and inducing personality in pre-trained language models . Preprint, arXiv:2206.07550

    arXiv 2023

  74. [77]

    John, Eileen M

    Oliver P. John, Eileen M. Donahue, and Robert L. Kentle. 1991. The big five inventory---versions 4a and 54. Technical report, University of California, Berkeley, Institute of Personality and Social Research

    1991

  75. [78]

    Bohao Li, Rui Wang, Guangzhi Wang, Yuying Ge, Yixiao Ge, and Ying Shan. 2023. https://arxiv.org/abs/2307.16125 Seed-bench: Benchmarking multimodal llms with generative comprehension . Preprint, arXiv:2307.16125

    Pith/arXiv arXiv 2023

  76. [79]

    Wenkai Li, Jiarui Liu, Andy Liu, Xuhui Zhou, Mona Diab, and Maarten Sap. 2025. https://arxiv.org/abs/2410.16491 Big5-chat: Shaping llm personalities through training on human-grounded data . Preprint, arXiv:2410.16491

    arXiv 2025

  77. [80]

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. 2023. https://arxiv.org/abs/2304.08485 Visual instruction tuning . Preprint, arXiv:2304.08485

    Pith/arXiv arXiv 2023

  78. [81]

    Yuan Liu, Haodong Duan, Yuanhan Zhang, Bo Li, Songyang Zhang, Wangbo Zhao, Yike Yuan, Jiaqi Wang, Conghui He, Ziwei Liu, Kai Chen, and Dahua Lin. 2024. https://arxiv.org/abs/2307.06281 Mmbench: Is your multi-modal model an all-around player? Preprint, arXiv:2307.06281

    Pith/arXiv arXiv 2024

  79. [82]

    Jiasen Lu, Dhruv Batra, Devi Parikh, and Stefan Lee. 2019. https://arxiv.org/abs/1908.02265 Vilbert: Pretraining task-agnostic visiolinguistic representations for vision-and-language tasks . Preprint, arXiv:1908.02265

    arXiv 2019

  80. [83]

    Pan Lu, Hritik Bansal, Tony Xia, Jiacheng Liu, Chunyuan Li, Hannaneh Hajishirzi, Hao Cheng, Kai-Wei Chang, Michel Galley, and Jianfeng Gao. 2024. https://arxiv.org/abs/2310.02255 Mathvista: Evaluating mathematical reasoning of foundation models in visual contexts . Preprint, arXiv:2310.02255

    Pith/arXiv arXiv 2024

Showing first 80 references.