Papers › TS-LLaVA: Constructing Visual Tokens through Thumbnail-and-Sampling for Training-Free...

TS-LLaVA: Constructing Visual Tokens through Thumbnail-and-Sampling for Training-Free Video Large Language Models

17 Nov 2024arXiv:2411.11066archive 2025-07-28

Tingyu Qu, Mingxiao Li, Tinne Tuytelaars, Marie-Francine Moens

Recent advances in multimodal Large Language Models (LLMs) have shown great success in understanding multi-modal contents. For video understanding tasks, training-based video LLMs are difficult to build due to the scarcity of high-quality, curated video-text paired data. In contrast, paired image-text data are much easier to obtain, and there is substantial similarity between images and videos. Consequently, extending image LLMs for video understanding tasks presents an appealing alternative. Developing effective strategies for compressing visual tokens from multiple frames is a promising way to leverage the powerful pre-trained image LLM. In this work, we explore the limitations of the existing compression strategies for building a training-free video LLM. The findings lead to our method TS-LLaVA, which constructs visual tokens through a Thumbnail-and-Sampling strategy. Given a video, we select few equidistant frames from all input frames to construct a Thumbnail image as a detailed visual cue, complemented by Sampled visual tokens from all input frames. Our method establishes the new state-of-the-art performance among training-free video LLMs on various benchmarks. Notably, our 34B model outperforms GPT-4V on the MVBench benchmark, and achieves performance comparable to the 72B training-based video LLM, Video-LLaMA2, on the challenging MLVU benchmark. Code is available at https://github.com/tingyu215/TS-LLaVA.

PaperPDFCodeCode Syntology ran

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, Syntology's MCP tool lists every function and class Syntology harvested from this paper and whether it ran (how to connect): get_harvested_code_for_paper(arxiv_id="2411.11066")

Code

Syntology Ran 5 of 11 code samples harvested from 1 repository linked to this paper; 6 have no recorded run. Of those that ran: 2 ran · our draft was wrong; 1 ran · fixture could not drive it; 2 ran with no contract checked.

By repository: official repository: 11 samples from 1 repository, 5 ran. The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

tingyu215/ts-llava officialmentioned in papermentioned on GitHubpytorchApache-2.0 report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

11 samples harvested; 5 ran; 0 honoured the contract we drafted; 6 have no recorded run. Read from Syntology's graph 2026-09-24; that is when this build read the record, not when the samples ran.

2ran · our draft was wrong
1ran · fixture could not drive it
2ran
6unverified

Licence: 0 of the 11 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from tingyu215/ts-llava. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself; “our draft was wrong” and “fixture could not drive it” are failures of Syntology's instrument, not of the code.

Each sample ends with its code_sha256, Syntology's identity for that exact code. An agent fetches the stored sample with Syntology's MCP tool get_code(code_sha256="…") (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples from a repository marked official are listed first. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of four reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; or the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar). Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

divide_to_patches tingyu215/ts-llava/llava/mm_utils.py official repository ran Apache-2.0 (permissive) · 7e03b180fa317c9a · report
get_chunk tingyu215/ts-llava/llava/eval/run_inference_benchmark_consistency.py official repository ran · our draft was wrong fingerprinted Apache-2.0 (permissive) · 42a46570620cd9fa · report
resize_and_pad_image tingyu215/ts-llava/llava/mm_utils.py official repository ran Apache-2.0 (permissive) · 468eedeba67f1b00 · report
select_best_resolution tingyu215/ts-llava/llava/mm_utils.py official repository ran · fixture could not drive it Apache-2.0 (permissive) · 3999ff487573f32c · report
split_list tingyu215/ts-llava/llava/eval/run_inference_benchmark_consistency.py official repository ran · our draft was wrong fingerprinted Apache-2.0 (permissive) · 076c252c52cbb161 · report
find_best_grid tingyu215/ts-llava/llava/model/llava_arch.py official repository unverified Apache-2.0 (permissive) · 075d76388f7066e2 · report
get_seq_frames tingyu215/ts-llava/llava/eval/model_utils.py official repository unverified Apache-2.0 (permissive) · 5764189adab1c8ff · report
pretty_print_semaphore tingyu215/ts-llava/llava/utils.py official repository unverified Apache-2.0 (permissive) · 37899f22fb191b37 · report
resize_images tingyu215/ts-llava/llava/model/llava_arch.py official repository unverified Apache-2.0 (permissive) · 35466d298191b5bd · report
uniformly_sample_frames tingyu215/ts-llava/llava/model/llava_arch.py official repository unverified Apache-2.0 (permissive) · 233b9a2ceb040553 · report
violates_moderation tingyu215/ts-llava/llava/utils.py official repository unverified Apache-2.0 (permissive) · f9939a84b9a65279 · report

Tasks

Video UnderstandingVideo-based Generative Performance BenchmarkingVideo-based Generative Performance Benchmarking (Consistency)Video-based Generative Performance Benchmarking (Contextual Understanding)Video-based Generative Performance Benchmarking (Correctness of Information)Video-based Generative Performance Benchmarking (Detail Orientation))Video-based Generative Performance Benchmarking (Temporal Understanding)Zero-Shot Video Question Answer

1 archive task tag without a task page not shown.

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Video-based Generative Performance Benchmarking VideoInstruct TS-LLaVA-34B mean 3.38 #3 of 23 Archive leaderboard report
Video-based Generative Performance Benchmarking (Consistency) VideoInstruct TS-LLaVA-34B gpt-score 3.69 #2 of 18 Archive leaderboard report
Video-based Generative Performance Benchmarking (Contextual Understanding) VideoInstruct TS-LLaVA-34B gpt-score 3.86 #3 of 18 Archive leaderboard report
Video-based Generative Performance Benchmarking (Correctness of Information) VideoInstruct TS-LLaVA-34B gpt-score 3.55 #3 of 18 Archive leaderboard report
Video-based Generative Performance Benchmarking (Detail Orientation)) VideoInstruct TS-LLaVA-34B gpt-score 3.03 #6 of 18 Archive leaderboard report
Video-based Generative Performance Benchmarking (Temporal Understanding) VideoInstruct TS-LLaVA-34B gpt-score 2.77 #5 of 18 Archive leaderboard report
Zero-Shot Video Question Answer ActivityNet-QA TS-LLaVA-34B Accuracy 58.9 #6 of 28 Archive leaderboard report
Zero-Shot Video Question Answer ActivityNet-QA TS-LLaVA-34B Confidence Score 3.5 #6 of 28 Archive leaderboard report
Zero-Shot Video Question Answer EgoSchema (subset) TS-LLaVA-34B Accuracy 57.8 #9 of 14 Archive leaderboard report
Zero-Shot Video Question Answer IntentQA TS-LLaVA-34B Accuracy 67.9 #3 of 13 Archive leaderboard report
Zero-Shot Video Question Answer MSRVTT-QA TS-LLaVA-34B Accuracy 66.2 #7 of 30 Archive leaderboard report
Zero-Shot Video Question Answer MSRVTT-QA TS-LLaVA-34B Confidence Score 3.6 #7 of 30 Archive leaderboard report
Zero-Shot Video Question Answer MSVD-QA TS-LLaVA-34B Accuracy 79.4 #8 of 28 Archive leaderboard report
Zero-Shot Video Question Answer MSVD-QA TS-LLaVA-34B Confidence Score 4.1 #8 of 28 Archive leaderboard report
Zero-Shot Video Question Answer MVBench TS-LLaVA-34B Accuracy 52.6 #1 of 2 Archive leaderboard report
Zero-Shot Video Question Answer NExT-QA TS-LLaVA-34B Accuracy 73.6 #5 of 27 Archive leaderboard report
Zero-Shot Video Question Answer TGIF-QA TS-LLaVA-34B Accuracy 81.0 #3 of 14 Archive leaderboard report
Zero-Shot Video Question Answer TGIF-QA TS-LLaVA-34B Confidence Score 4.2 #3 of 14 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections