First author of CapMem, a caption-based episodic memory benchmark probing vision-language models on egocentric video
1
papers here
1
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-21) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
No output data is available yet. This axis would measure the volume and consistency of research production — number of publications, venues, and authorship roles over time — none of which can be computed from the single catalogued 2026 paper.
Impact—
Impact data is not yet available. This axis would measure citation-based influence — citations, adoption, and downstream use of the CapMem benchmark by other teams — which cannot be scored while the data contains no citation record for the 2026 paper.
Focus100/100
The focus score is 100/100, computed from the field distribution of catalogued output: the single reviewed paper sits entirely in computer science (computer vision / multimodal AI), giving a one-field distribution with weight 1. Every catalogued contribution — benchmark construction, caption-versus-VideoQA evaluation, VLM memory probing — belongs to this one area, so measured topical concentration is maximal, though it rests on a single 2026 paper.
Collaboration44/100
The collaboration score is 44/100, based on 8 distinct coauthors recorded for the catalogued work. The network currently sits entirely within the CapMem orbit — frequent collaborators include Yiqiao Xie, Yukai Huang, Zhaokai Wang, Weitong Cai, Guangwen Feng, Jifei Song, Zhensong Zhang and Hang Zhang — indicating a moderate, team-based footprint rather than a broad cross-institution web; more catalogued papers would clarify its true breadth.
Activity—
Activity data is insufficient. This axis would measure recent publishing cadence — outputs per year and continuity of research tempo — but only one 2026 entry exists, so current productivity level and momentum cannot be assessed honestly.
Reproducibility55/100
Reproducibility scores 55/100, computed as the mean (reproMean 55) over one scored paper (reproPapers 1): CapMem, whose reproducibility readiness Evolve rates as medium. As a benchmark it specifies an evaluation across 12 VLMs, which aids replication, but the verdict that controlled gains are Qwen-only and partly null means cross-model replication needs care; the score is provisional given the single-paper base.
Research summary
Dingli Liang is an early-stage researcher profile in computer vision and multimodal AI, centered so far on evaluation methodology for vision-language models (VLMs). The Evolve catalog holds one reviewed paper, CapMem: A Benchmark for Caption-Based Episodic Memory in Egocentric Video (arXiv, 2026), with Liang as first author. The paper proposes probing VLMs' episodic memory on egocentric video through captions rather than direct video question answering; Evolve's verdict calls it a credible benchmark, finding that captions beat direct VideoQA on long egocentric videos for most of 12 VLMs tested, while noting that controlled gains are Qwen-only and partly null. Bibliometrically the profile is maximally focused: 100% of catalogued output sits in a single field — computer science (computer vision / multimodal AI) — and the recorded network spans 8 distinct coauthors, including frequent collaborators Yiqiao Xie, Yukai Huang, Zhaokai Wang, Weitong Cai, Guangwen Feng, Jifei Song, Zhensong Zhang and Hang Zhang. Axis scores exist for focus (100), reproducibility (55, from one paper) and collaboration (44); output, impact and activity lack sufficient data. With no institution, country, or ORCID attached, this profile is name-resolved from public bibliographic data.
Collaboration pattern
All recorded collaboration sits on the 2026 CapMem paper: 8 distinct coauthors, all listed as frequent collaborators — Yiqiao Xie, Yukai Huang, Zhaokai Wang, Weitong Cai, Guangwen Feng, Jifei Song, Zhensong Zhang and Hang Zhang. The pattern suggests one core team producing the benchmark together, with Liang in the first-author position; the collaboration score of 44 reflects a moderate, single-project network whose breadth cannot yet be judged from one catalogued paper.
Research keywords
Highlights
Captions beat direct VideoQA on long egocentric videos
A credible benchmark that is honest about its limits
First-author benchmark anchoring an early-stage profile
Timeline
- 2026CapMem, a benchmark for caption-based episodic memory in egocentric video, is posted on arXiv with Dingli Liang as first author.
- 2026Evolve reviews CapMem, judging it a credible benchmark: captions beat direct VideoQA on long egocentric videos for most of 12 VLMs, while controlled gains are Qwen-only and partly null.
- 2026The paper enters the Evolve catalog, establishing a single-field (computer vision / multimodal AI) profile with 8 recorded coauthors and a medium reproducibility rating.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometric scores rest on a very thin base: one catalogued paper (2026), so focus=100, reproducibility=55 and collaboration=44 should be read as provisional.
- Output, impact and activity axes are unscored because no publication-volume or citation data is available in the input.
- No institution, country, or ORCID is attached; the profile is name-resolved from public bibliographic data and may merge or split namesakes.
- Paper-level facts come from Evolve's review of an arXiv preprint; peer-reviewed venue status is not recorded.