Jayanth Srinivasa

Cisco Systems (United States) · United States

Unclaimed · AI-structured profileORCID OpenAlex

Cisco researcher advancing LLMs for code — generation, editing, evaluation, and reliable tool-calling agents

papers here

reviewed by Evolve

first-author

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—

Output scoring is unavailable: this snapshot lacks the fields required to compute the dimension. It would measure the volume, consistency, and growth of publications over time; recorded year counts and 53 recent works indicate scale but yield no score.

Impact—

Impact scoring is unavailable for this profile. The axis would measure citation uptake, influence on follow-up work, and recognition within the field. The recorded h-index of 6 (i10-index 6) points to an early-stage footprint, but no impact score is computed.

Focus0/100

The recorded focus score is 0, indicating a broad rather than concentrated footprint. The field distribution is an even 50/50 split between NLP/software engineering (LLMs for code) and LLM-based code generation — closely related areas — while topic mass spreads across five labels: topic modeling (29), NLP techniques (20), AI-based problem solving and planning (12), adversarial robustness (9), and multimodal applications (9).

Collaboration54/100

Collaboration scores 54 on the 0-100 scale. The recent three-year window records 53 works with 14 distinct coauthors, a moderate co-authoring density. The recurring list spans twelve names — among them Junyi Jessy Li, Milos Gligoric, and Chitta Baral alongside Aditya Thimmaiah, Amir Saeidi, Ali Payani, and Gaowen Liu — suggesting a stable core team plus rotating collaborators rather than one-off pairings.

Activity100/100

Activity scores 100, the top of the scale. Recorded works per year rise from 1 in 2021 and 1 in 2022 to 7 in 2023, 13 in 2024, and 20 in both 2025 and 2026 — a steep, sustained ramp. The 53-work recent three-year total confirms a high-tempo cadence that is still climbing or holding at peak.

Reproducibility78/100

Reproducibility scores 78 (reproMean) across 2 scored papers. Within the catalog, the exception-retrofitting work and the PrEx benchmark are rated high readiness, while CROCODIL is rated medium — its gains lack statistical tests and use a single base model. The pattern suggests attention to reusable benchmarks, harnesses, and multi-model evidence, with room to tighten statistical rigor.

Research summary

Jayanth Srinivasa is a researcher at Cisco Systems (United States) whose recent work centers on large language models for code: generating, editing, and evaluating programs, and extending them toward reliable long-horizon tool-calling agents. This profile is name-resolved from public bibliographic data (OpenAlex). His recorded output accelerates sharply: one work each in 2021 and 2022, then 7 (2023), 13 (2024), and 20 in each of 2025 and 2026 — 53 works in the recent three-year window with 14 distinct coauthors. The h-index and i10-index are both 6, an early-stage citation footprint, and the output and impact dimensions lack sufficient data to score. Topic mass concentrates in topic modeling (29), NLP techniques (20), AI-based planning (12), adversarial robustness (9), and multimodal machine learning (9), while the field split is even between NLP/software engineering for code and LLM-based code generation. Evolve-reviewed papers include retrofitting Java code with exception behavior derived from tests (consistent large gains across 5 models, though roughly 25% of test-passing code differs from intent), CROCODIL's GRPO reward that halves Olmo3 7B edits while preserving success, and the PrEx benchmark showing open-source LLMs answer from priors rather than supplied semantics. Reproducibility averages 78 across 2 scored papers, with two rated high readiness.

Collaboration pattern

Collaboration is mid-range (score 54): 14 distinct coauthors across 53 recent three-year works. The recurring roster lists twelve names — Junyi Jessy Li, Milos Gligoric, Aditya Thimmaiah, Linghan Zhong, Amir Saeidi, Zehua Zhang, Rishitosh Singh, Naman Ahuja, Vivek Gupta, Ali Payani, Gaowen Liu, and Chitta Baral — pointing to a stable core team plus rotating partners rather than one-off pairings.

Research keywords

LLM-based code generationLLM code editingException handling retrofittingTest-driven developmentStatic and dynamic program analysisOperational semanticsLLM evaluation and benchmarkingReinforcement learning (GRPO) and reward designLong-horizon tool-calling agentsTopic modelingAdversarial robustnessMultimodal machine learning

Highlights

Retrofitting exception behavior into code with LLMs
Evolve's review notes consistent large LLM gains across 5 models when retrofitting code to support exceptional behavior from tests, while cautioning that roughly 25% of test-passing code still differs from intent.
CROCODIL curbs excessive cross-model code edits
LLMs tend to over-edit foreign code; CROCODIL's GRPO-based reward halves Olmo3 7B edits while preserving success. The review credits a good corpus but flags missing statistical tests and a single base model.
PrEx: open-source LLMs lean on priors, not supplied semantics
Predicting program exit codes with LLMs and programming language semantics, the benchmark finds sharp drops under semantic shifts — evidence models answer from priors. The review notes means-only statistics and no heuristic baseline.
Critique-aware supervision for long-horizon tool-calling agents
CAST trains reliable long-horizon tool-calling agents using critique-aware supervision. This catalog entry has not yet been reviewed by Evolve, so no verdict, keywords, or reproducibility rating is available.

Timeline

2021First recorded work year in the OpenAlex snapshot, with 1 work.
2022Output remained minimal at 1 recorded work.
2023Recorded works rose to 7, marking the start of a sustained ramp.
202413 recorded works, nearly doubling the previous year.
202520 recorded works, reaching the snapshot's high plateau.
202620 recorded works so far, including four catalog papers: exception retrofitting, CROCODIL, PrEx, and CAST.

Papers in the Evolve catalog

2026-09-09
Retrofitting Code Using LLMs to Support Exceptional Behavior
2026-09-03
CROCODIL: Cross-Model Code Editing with LLMs
2026-09-01
Predicting Program Exit Code with LLMs and Programming Language Semantics

Frequent co-authors

Junyi Jessy Li×3 Milos Gligoric×3 Aditya Thimmaiah×2 Linghan Zhong×2 Amir Saeidi×1 Zehua Zhang×1 Rishitosh Singh×1 Naman Ahuja×1 Vivek Gupta×1 Ali Payani×1 Gaowen Liu×1 Chitta Baral×1

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 figures (h-index, i10-index, counts by year, topics, coauthors) come from an OpenAlex snapshot and may undercount non-indexed output.
Output and impact dimension scores are unavailable (dimMissing); no citation-impact scoring is attempted here.
The reproducibility mean (78) covers only 2 scored papers; the CAST catalog entry has no Evolve verdict, keywords, or slug yet.
Profile is name-resolved from public bibliographic data; biography beyond this snapshot and the reviewed catalog is not covered.