SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2412.05185 · 2024

LinVT: Empower Your Image-level Large Language Model to Understand Videos

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 12 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
gls0425/linvt canonical 1 of 9
copy not recorded — 3 of 3
FunctionStatusWhere it lives
evaluate_exact_match_accuracy Ran this paper's copy was not recorded; identical code first harvested from brandon3964/multimodal-task-vector
pointer only · get_code("e159a001bd4cbfaa")
evaluate_relaxed_accuracy Ran this paper's copy was not recorded; identical code first harvested from brandon3964/multimodal-task-vector
pointer only · get_code("4ffcad52354b4806")
find_closest_aspect_ratio Ran gls0425/linvt/streamlit_demo/model_worker.py
pointer only (licence: NONE) · get_code("bd77f5f8067f18e9")
relaxed_correctness Ran this paper's copy was not recorded; identical code first harvested from brandon3964/multimodal-task-vector
pointer only · get_code("b3c8680ebd8739af")
allreduce_hook Not yet run gls0425/linvt/classification/ddp_hooks.py
pointer only (licence: NONE) · get_code("bb1a26fe42d7075c")
bf16_compress_hook Not yet run gls0425/linvt/classification/ddp_hooks.py
pointer only (licence: NONE) · get_code("21bd5be56c8f79d0")
build_scheduler Not yet run gls0425/linvt/classification/lr_scheduler.py
pointer only (licence: NONE) · get_code("8c4fa54e6e3b7fb7")
create_logger Not yet run gls0425/linvt/classification/logger.py
pointer only (licence: NONE) · get_code("45fcbc909d30fb3b")
fp16_compress_hook Not yet run gls0425/linvt/classification/ddp_hooks.py
pointer only (licence: NONE) · get_code("26bfffb625dfe211")
get_config Not yet run gls0425/linvt/classification/config.py
pointer only (licence: NONE) · get_code("e452206cf1a59c19")
load_image_from_base64 Not yet run gls0425/linvt/streamlit_demo/model_worker.py
pointer only (licence: NONE) · get_code("0dde2e782959c0bd")
sa_flops Not yet run gls0425/linvt/classification/gflops.py
pointer only (licence: NONE) · get_code("2dcfa48e398dddca")

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 Language Models (LLMs) have been widely used in various tasks, motivating us to develop an LLM-based assistant for videos. Instead of training from scratch, we propose a module to transform arbitrary well-trained image-based LLMs into video-LLMs (after being trained on video data). To better adapt image-LLMs for processing videos, we introduce two design principles: linear transformation to preserve the original visual-language alignment and representative information condensation from redundant video content. Guided by these principles, we propose a plug-and-play Linear Video Tokenizer(LinVT), which enables existing image-LLMs to understand videos. We benchmark LinVT with six recent visual LLMs: Aquila, Blip-3, InternVL2, Mipha, Molmo and Qwen2-VL, showcasing the high compatibility of LinVT. LinVT-based LLMs achieve state-of-the-art performance across various video benchmarks, illustrating the effectiveness of LinVT in multi-modal video understanding.

For agents

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

get_harvested_code_for_paper("2412.05185")
get_code_for_paper("2412.05185")
have("2412.05185")

Connect an agent — have() is free.