Senior researcher on efficient vision-language-action models for closed-loop autonomous driving
2
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-28) 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—
Output data is insufficient: the OpenAlex record reports h-index and i10-index of 0 and just 3 recent works, so total publication volume cannot be scored. This axis would measure cumulative output and productivity over time.
Impact—
Citation-based impact cannot be scored: no citation counts or citation distribution are available for this identity, whose zero index values suggest incomplete author disambiguation. This axis would measure citation accumulation and influence on later work.
Focus100/100
Focus scores 100/100: the field distribution assigns the catalog record entirely to autonomous driving (computer vision/robotics), and both reviewed 2026 papers are driving VLAs. Topic labels concentrate on advanced neural network and multimodal ML applications (2 each), reinforcing a tightly scoped line around end-to-end, closed-loop driving.
Collaboration36/100
Collaboration scores 36/100, based on 5 distinct recent coauthors across 3 works in the last three years — small, stable teams rather than broad networks. Recurring names include Alexey Zakharov, Kemal Oksuz, Alexandru Buburuzan and Yuhan Yao, and last-author placement on both catalog papers indicates a senior supervisory role.
Activity40/100
Activity scores 40/100 from recent3yWorks = 3, with countsByYear showing 1 work in 2025 and 2 in 2026. Output is recent and clustered in the current year, but the short recorded window and partial identity mean the cadence estimate rests on very few datapoints.
Reproducibility55/100
Reproducibility scores 55/100, a mean taken over 1 evaluated paper. FIVE-VLA carries medium repro-readiness: Evolve's review credits thorough ablations but flags thin attribution and statistics, and its 641M-parameter closed-loop setup implies non-trivial compute for independent replication.
Research summary
Puneet K. Dokania is profiled here from OpenAlex bibliographic records and two 2026 arXiv preprints reviewed by Evolve, on both of which he appears as last author. His recent catalog output centers on vision-language-action (VLA) models for autonomous driving: FIVE-VLA, a 641M-parameter driving VLA combining 98 vision tokens with a recurrent action memory that reports state-of-the-art results on the Bench2Drive closed-loop benchmark, and Less Language, More Latents, which targets annotation-efficient VLAs for driving. Together they indicate a research line that reduces language supervision and computation in end-to-end driving stacks while keeping closed-loop evaluation central. The OpenAlex snapshot records 3 works in the last three years (1 in 2025, 2 in 2026), 5 distinct recent coauthors, and a 100/100 focus score, consistent with tight concentration on driving. Reported h- and i10-index values of 0 signal an incomplete identity record rather than measured absence of impact, and topic labels extend to generative modelling and gene-regulatory-network analysis, which may reflect name-ambiguity noise.
Collaboration pattern
The recorded pattern is small-team: 5 distinct coauthors over 3 works in the last three years, with Alexey Zakharov, Kemal Oksuz, Alexandru Buburuzan and Yuhan Yao recurring most often. Both catalogued 2026 papers place Dokania as last author (not flagged corresponding), consistent with a senior supervisory role in a compact, driving-focused group.
Research keywords
Highlights
FIVE-VLA: compact 641M-parameter driving VLA claims closed-loop SOTA
Annotation-efficient VLAs: less language, more latents
This 2026 arXiv preprint, with Dokania as last author, pushes driving VLAs toward lower annotation cost by leaning on latent representations rather than heavy language supervision, complementing the efficiency agenda of FIVE-VLA.
Consistent senior (last-author) role in 2026 driving papers
On both catalogued 2026 papers Dokania appears as last author without corresponding-author flagging, a placement that typically denotes lab leadership and supervision of the efficient-driving research line.
Timeline
- 2025One work recorded in the OpenAlex countsByYear; no title is available in this snapshot.
- 2026Less Language, More Latents: Annotation-Efficient VLAs for Driving (arXiv), last author.
- 2026FIVE-VLA: recurrent-action-memory driving VLA reviewed by Evolve; medium repro-readiness, claims Bench2Drive closed-loop SOTA.
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.
- Bibliometrics come from a single OpenAlex identity (identity = openalex); reported h-index and i10-index are 0, indicating incomplete author disambiguation rather than measured zero impact.
- Institution, country and ORCID are absent from the input data, so no affiliation or biographical context is asserted in this profile.
- Topic labels include gene-regulatory-network analysis and GAN image synthesis alongside driving topics, which may reflect name-ambiguity noise in the OpenAlex record.
- Only 2 papers have been reviewed by Evolve; the reproducibility score (55) rests on a single evaluated paper (reproPapers = 1).