Harvard researcher spanning LLM reasoning, feed-forward 3D Gaussian Splatting, and spoken dialogue state tracking
3
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) 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—
Not yet scored: the snapshot contains no publication counts, venue list, or yearly output totals. This axis would measure the volume, consistency, and venue quality of the researcher's published record over time.
Impact—
Not yet scored: no citation counts, h-index, or venue-level influence data are available. This axis would measure how often and how widely the reviewed work is cited and taken up by other researchers.
Focus0/100
The focus score is 0/100, derived from a field distribution that splits evenly — 0.5 natural language processing / LLM reasoning and 0.5 computer vision (3D scene reconstruction / novel view synthesis) — with a further spoken-dialogue strand. This profiles a deliberately broad, two-track agenda rather than a concentrated specialty.
Collaboration54/100
Collaboration scores 54/100 from 14 distinct coauthors in the snapshot, with 12 recurring partners recorded, including Kailong Fan, Anqi Pu, Yichen Wu, Wanhua Li, and Fangneng Zhan. Authorship sits mid-list on the reasoning and 3DGS papers and last on AVERT, with no corresponding-author role flagged — a deep-contributor position within mid-sized teams.
Activity—
Not yet scored: no multi-year publication timeline exists beyond three 2026 preprints, so publishing cadence cannot be assessed. This axis would measure recent output rate and research tempo across successive years.
Reproducibility55/100
Reproducibility averages 55/100 over two assessed papers (reproPapers = 2). Readiness ratings are medium for the medical-QA reasoning paper and for AVSplat, and low for AVERT; Evolve further flags thin statistical support on one paper and a missing assist-only ablation on the other.
Research summary
Hanspeter Pfister is a Harvard University researcher whose cataloged work spans two clearly separated lines: natural language processing — reasoning and dialogue — and computer vision, specifically 3D scene reconstruction with feed-forward Gaussian Splatting. The Evolve-reviewed snapshot covers three 2026 arXiv preprints. In the reasoning line, 'Rewarding Reasoning, Not Answers' diagnoses why test-time reinforcement learning collapses on medical multiple-choice QA, attributing the failure to small answer spaces, with PROSE reported effective, though Evolve notes the statistical support remains thin. In the vision line, 'AVSplat' proposes AVP+OVF to counter dense-view degradation in pose-free feed-forward 3DGS. A third paper, 'AVERT', adds audio-verified adjudication to spoken dialogue state tracking, lifting SpokenWOZ joint goal accuracy from 38.34 to 40.13. Across these papers the researcher appears as a middle author twice and last author once, with no corresponding-author role recorded. The radar shows a collaboration score of 54 with 14 distinct coauthors, a focus score of 0 reflecting an even 50/50 field split, and a reproducibility score of 55 averaged over two assessed papers. Output, impact, and activity metrics are not yet available.
Collaboration pattern
The snapshot records 14 distinct coauthors and 12 recurring collaborators, with Kailong Fan, Anqi Pu, Yichen Wu, Wanhua Li, and Fangneng Zhan among the most frequent; collaboration scores 54/100. Authorship sits mid-list on the LLM-reasoning and 3DGS papers and last on AVERT, with no corresponding-author role recorded — a deep-contributor position inside multi-person teams spanning two fields.
Research keywords
Highlights
Diagnosing test-time RL collapse on medical QA
AVSplat targets dense-view degradation in pose-free feed-forward 3DGS
Audio-verified adjudication lifts SpokenWOZ accuracy
Timeline
- 2026Middle author on 'Rewarding Reasoning, Not Answers', fixing and bounding test-time reinforcement learning for medical QA.
- 2026Middle author on 'AVSplat', targeting dense-view degradation in pose-free feed-forward 3D Gaussian Splatting.
- 2026Last author on 'AVERT', adding audio-verified adjudication to spoken dialogue state tracking.
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.
- Radar scores and bibliometrics come from a limited public bibliometric snapshot; only the three Evolve-reviewed 2026 arXiv preprints inform this profile.
- Identity is name-resolved from public bibliographic data and the provided ORCID (0000-0002-3620-2582); records under this name could span multiple people.
- Output, impact, and activity axes lack underlying data (no citation counts or publication totals); no values are estimated.
- The reproducibility mean (55) covers only 2 of the 3 reviewed papers, and the peer-review status of these arXiv preprints is unknown.