Computational neuroscientist at Bristol modelling cerebellar motor learning, now extending into LLM-agent benchmarking
2
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
23
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—
This axis measures cumulative research output — total works and their growth over time. The snapshot supplies year-by-year counts from 2007 onward but no complete works total, so no score can be computed; the axis is honestly left as data-insufficient rather than estimated.
Impact—
This axis measures citation-based influence. The record lists an h-index of 23 and an i10-index of 33, but the citation aggregates required to score this axis are not available in the snapshot, so it is reported as data-insufficient rather than scored.
Focus100/100
The focus score of 100 reflects a tightly defined agenda. Topic assignments concentrate in neuroscience — "Neural dynamics and brain function" (71), "Memory and Neural Mechanisms" (55) and "Neuroscience and Neuropharmacology Research" (40) — and the field distribution lists a single field, computational neuroscience / motor control. "Neural Networks and Applications" (21) and "Advanced Memory and Neural Computing" (11) form a smaller machine-learning strand that the 2026 CivBench benchmark extends.
Collaboration41/100
The collaboration score of 41 sits in the mid-range. Over the last three years the record shows 37 works with only 7 distinct coauthors, indicating compact, small-team authorship rather than large consortia. Named frequent coauthors include Jakob Nicolaus Foerster, Cláudia Soares, Ana Carolina Filipe, Austin Tudor David Andrews, Liam Wilkinson, Jamie Heagerty and Harry Coppock.
Activity100/100
The activity score of 100 matches the raw record: 37 works in the last three years, with annual counts climbing from 9 in 2024 to 13 in 2025 and 15 in 2026 — the highest annual total since the record began in 2007. Output has been essentially continuous for two decades, with earlier peaks of 12 works in 2017 and 10 in 2019.
Reproducibility100/100
The reproducibility score of 100 reflects a mean readiness of 100, but it rests on a single Evolve-assessed paper. That paper — the cortico-cerebellar motor-learning model — is rated high reproducibility readiness, consistent with a code-and-model-driven RNN study. With reproPapers at 1, this axis should be read as a promising single data point, not a portfolio-wide guarantee.
Research summary
Rui Ponte Costa is a computational neuroscientist at the University of Bristol whose OpenAlex-sourced record spans 2007–2026, with an h-index of 23 and an i10-index of 33. The publication profile is strongly neuroscience-centred: topic assignments are led by "Neural dynamics and brain function" (71 works), "Memory and Neural Mechanisms" (55) and "Neuroscience and Neuropharmacology Research" (40), with smaller strands in "Neural Networks and Applications" (21) and "Advanced Memory and Neural Computing" (11). Activity is currently at its highest recorded level: 37 works in the last three years, with annual counts rising from 9 (2024) to 13 (2025) and 15 (2026), yielding an activity score of 100. Recent work shows two faces: a cerebellum-inspired recurrent-network model of error- and prediction-driven motor learning in the cortico-cerebellar loop, rated by Evolve at high reproducibility readiness, and CivBench, a long-horizon benchmark for tool-mediated LLM agents in Civilization VI built on the Model Context Protocol. Recent collaboration is compact — 37 works across only 7 distinct coauthors — giving a mid-range collaboration score of 41. The radar cannot currently score output or impact, so those axes are left unassessed rather than estimated.
Collaboration pattern
Collaboration is compact: the last three years yield 37 works across only 7 distinct coauthors, matching the mid-range collaboration score of 41. Recurring partners include Jakob Nicolaus Foerster, Cláudia Soares, Ana Carolina Filipe, Austin Tudor David Andrews, Liam Wilkinson, Jamie Heagerty and Harry Coppock, spanning the neuroscience modelling line and the newer agent-benchmark work. The pattern suggests small, hands-on teams rather than large multi-institution consortia.
Research keywords
Highlights
Cerebellar-loop model of error- and prediction-driven motor learning
CivBench: long-horizon benchmark for tool-mediated LLM agents
Timeline
- 2007Earliest works appear in the OpenAlex record (1 publication).
- 2013Annual output reaches 7 works, the highest level to that point.
- 2017A first major peak: 12 works published in a single year.
- 2020Output dips to 2 works, the lowest annual count since 2007.
- 2026Record-high annual total of 15 works (after 13 in 2025), including two Evolve-catalogued arXiv preprints on cortico-cerebellar motor learning and the CivBench agent benchmark.
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 indicators come from a single OpenAlex-sourced snapshot tied to ORCID 0000-0003-2595-2027; no other databases are aggregated.
- The radar could not compute the output and impact axes from this snapshot, so no scores are offered for them.
- Paper-level assessments rest on only two Evolve-reviewed papers (both 2026 arXiv preprints); coverage of the wider portfolio is thin.
- The reproducibility mean of 100 is based on a single assessed paper (reproPapers: 1).