UNC Chapel Hill robotics researcher working on VLA policy robustness and long-horizon manipulation evaluation.
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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 metrics — total publication count, venue mix, and yearly volume — are not yet available in the underlying bibliometric data, so this axis is left unscored. It would measure the scale and consistency of Szafir's publishing record.
Impact—
Impact signals such as citation counts, h-index, and downstream adoption are absent from the current snapshot, so no score is assigned. This axis would measure how widely the VLA robustness and evaluation-harness work is taken up by other research groups.
Focus0/100
The focus axis scores 0 because the raw field distribution splits exactly 50/50 between robot learning (vision-language-action models, multimodal robotics) and robotics (robot learning and manipulation evaluation). The two 2026 papers indeed attack different problems — one hardens VLA policies against modality entanglement, the other builds an autonomous reset harness for evaluation — so the current record spans two adjacent subfields rather than a single concentrated line.
Collaboration46/100
Szafir's collaboration axis scores 46, computed from a pool of 9 distinct coauthors across the two cataloged papers. The recurring co-author names — Gedas Bertasius, Yue Yang, Diego Romeres, Chiori Hori, Siddarth Jain, Jing Jiang, Xinkai Jiang, and Rudolf Lioutikov — span several groups, and both 2026 papers place him in a middle, non-corresponding authorship position, indicating steady project-based teamwork rather than solo work or a very large network.
Activity—
Recent-activity measures such as publication cadence and year-over-year output cannot yet be computed from the data on hand. This axis would track publishing rhythm over time; only the two 2026 arXiv papers are currently observable.
Reproducibility38/100
Reproducibility scores 38, the mean readiness rating across 2 papers. The Evidence-Gated Regularization paper is rated medium — notable simulated and real-robot gains with honest statistics, but key ablations and baselines are missing. The HALTER reset-harness paper is rated low, with solid real-robot gains but thin small-N statistics. Publishing code, seeds, and complete baseline comparisons would be the clearest route to raising this score.
Research summary
Daniel J. Szafir is a robotics researcher at the University of North Carolina at Chapel Hill whose recent work bridges vision-language-action (VLA) robot learning and rigorous manipulation evaluation. The Evolve catalog currently holds two of his 2026 arXiv papers, in both of which he is a middle author. 'Sensing Which Modality Matters' introduces Evidence-Gated Regularization (EGR), which gates consistency losses by per-sensor evidence to combat modality entanglement; the paper reports notable simulated and real-robot gains with honest statistics, though key ablations and baselines are still missing. 'From Rollout to Reset' presents HALTER, a graph-based harness whose LLM plans atomic-skill resets over scene graphs, enabling autonomous long-horizon manipulation evaluation on real robots; gains are solid but small-N statistics remain thin. His field distribution splits evenly between robot learning and manipulation-evaluation robotics (0.5/0.5). Radar scores put collaboration at 46 with 9 distinct coauthors and reproducibility at 38 (mean across 2 papers), while output, impact, and activity metrics are not yet available. This early-stage, name-resolved profile draws only on public bibliographic data.
Collaboration pattern
Collaboration is moderate: radar score 46, drawn from 9 distinct coauthors across the two 2026 papers, in both of which Szafir is a middle, non-corresponding author. The recurring co-author list includes Gedas Bertasius, Yue Yang, Diego Romeres, Chiori Hori, Siddarth Jain, Jing Jiang, Xinkai Jiang, and Rudolf Lioutikov, suggesting teamwork across several groups and projects rather than within one large standing lab.
Research keywords
Highlights
Evidence-Gated Regularization hardens VLA policies against modality entanglement
HALTER enables autonomous long-horizon manipulation evaluation
Timeline
- 2026Published 'Sensing Which Modality Matters: Evidence-Gated Regularization for Robust VLA Policies' on arXiv (middle author).
- 2026Published 'From Rollout to Reset: A Graph-Based Harness for Autonomous Long-Horizon Manipulation Evaluation' on arXiv (middle author).
- 2026Cataloged output splits evenly across his two fields — VLA robot learning and manipulation-evaluation robotics (0.5/0.5).
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 come from public bibliometric data only; the output, impact, and activity axes lack sufficient data and are unscored.
- Scope is limited to two 2026 arXiv papers reviewed by Evolve; earlier publications, citation counts, and h-index are out of scope.
- Both papers list the researcher as a middle, non-corresponding author, so individual contribution cannot be precisely attributed.
- Identity is name-resolved from public bibliographic data (ORCID 0000-0003-1848-7884); country and affiliation details are recorded as provided.