SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2505.24449 · ICLR · 2025

When Large Multimodal Models Confront Evolving Knowledge: Challenges and Explorations

Qing Li, Bin Li, Zilong Zheng, Yuchen Ren, Zhi Gao, Lei Liu, Kailin Jiang, Yuntao Du, Yukai Ding, Ning Jiang

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 6 functions out of this paper's own repositories and ran 4 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
pjlab-sys4nlp/llama-moe reimplementation 3 of 3
deepseek-ai/DeepSeek-MoE reimplementation 1 of 1
copy not recorded — 0 of 2
FunctionStatusWhere it lives
softmax Ran pjlab-sys4nlp/llama-moe/smoe/entrypoint/eval/eval_mmlu_moe_0.py
code served (permissive licence) · get_code("19f01570b0a18e2b")
build_instruction_prompt Ran deepseek-ai/DeepSeek-MoE/finetune/finetune.py
code served (permissive licence) · get_code("765ee58c7d78fb1b")
format_example Ran pjlab-sys4nlp/llama-moe/smoe/entrypoint/eval/eval_mmlu_moe_0.py
code served (permissive licence) · get_code("cd763eaf1ac287e7")
format_subject Ran pjlab-sys4nlp/llama-moe/smoe/entrypoint/eval/eval_mmlu_moe_0.py
code served (permissive licence) · get_code("6ab745408cb8648b")
save_image_to_local Not yet run this paper's copy was not recorded; identical code first harvested from PKU-YuanGroup/MoE-LLaVA
pointer only · get_code("e2e5c3a95a6aebb4")
save_video_to_local Not yet run this paper's copy was not recorded; identical code first harvested from PKU-YuanGroup/MoE-LLaVA
pointer only · get_code("22930d58d32e2d08")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Large Multimodal Models (LMMs) store vast amounts of pretrained knowledge but struggle to remain aligned with real-world updates, making it difficult to avoid capability degradation when acquiring evolving knowledge. Furthermore, most current work focuses on exploring static textual knowledge injection, neglecting dynamic multimodal evolving knowledge injection, leaving the potential of LMMs for multimodal knowledge injection as an open question. To address this, we first propose a pipeline to construct MMEVOKE, a benchmark for evaluating LMMs' ability in multimodal evolving knowledge injection. MMEVOKE contains 9,422 samples spanning 159 subtypes. Then, based on extensive experiments with MMEVOKE, we reveal challenges such as poor injection performance and capability degradation in existing knowledge injection methods through knowledge injection tests and general capability tests. Finally, to tackle these challenges, we introduce knowledge augmentation and knowledge retention methods, finding that knowledge-aware augmentation strengthens knowledge injection performance, and that Data Replay and MoE methods effectively mitigate capability degradation. Project Page: https://evoke-lmm.github.io/ * Equal contribution. † Corresponding author.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2505.24449")
get_code_for_paper("2505.24449")
have("2505.24449")

Connect an agent — have() is free.