# Agentic System

8 related articles

Subscribe to updates

arXiv Nov 7, 2026
Radiology agents limited to cleared tools recover 29% of actionable findings when planning from the indication alone
A radiology agentic framework separates a large-language-model orchestrator from a modular tool interface and uses 1,989 finding classes consolidated from FDA-cleared radiology AI devices as its tool catalog, simulating structured report generation on the PARROT fictional-report corpus. Planning from the clinical indication alone, actionable-finding recall is 0.290 with the cleared catalog and 0.437 with an unlimited catalog; when the planner reads the target report, these rise to 0.426 and 0.740. Tool outputs are simulated from the reference report, so the comparison isolates tool coverage and orchestration rather than tool error.
A radiology agentic framework separates a large-language-model orchestrator from a modular tool interface and uses 1,989 finding classes consolidated from FDA-cleared radiology AI devices as its tool catalog, simulating structured report generation on the PARROT fictional-report corpus. Planning from the clinical indication alone, actionable-finding recall is 0.290 with the cleared catalog and 0.437 with an unlimited catalog; when the planner reads the target report, these rise to 0.426 and 0.740. Tool outputs are simulated from the reference report, so the comparison isolates tool coverage and orchestration rather than tool error.
Aubin Charley（First author）
Agentic System Radiology Report Generation ReAct loop
medRxiv Oct 14, 2026
An agent emulates an elderly clopidogrel trial: 53 of 62 SQL checks met, calibrated risk ratio point estimate rises to 1.274
TrialCalibre1 links a clinical question, study design, and executable analysis into one reviewable chain, covering reference-trial replication, direct target-trial emulation, and BenchExCal, which joins benchmarking, extension, and calibration. On synthetic PCORnet-style claims it used ASCET as benchmark and extended the clopidogrel-versus-aspirin comparison to ages 81–90: 6,069 benchmark and 1,602 extension patients, extension risk ratio 0.848 (95% CI 0.569–1.184), calibrated to a point estimate of 1.274 (0.439–3.700) with a sensitivity interval still spanning 1. Two reviewers rated 53/62 and 50/62 assessable SQL checks met and 12/13 R checks met, with protocol and proxy disagreements remaining.
TrialCalibre1 links a clinical question, study design, and executable analysis into one reviewable chain, covering reference-trial replication, direct target-trial emulation, and BenchExCal, which joins benchmarking, extension, and calibration. On synthetic PCORnet-style claims it used ASCET as benchmark and extended the clopidogrel-versus-aspirin comparison to ages 81–90: 6,069 benchmark and 1,602 extension patients, extension risk ratio 0.848 (95% CI 0.569–1.184), calibrated to a point estimate of 1.274 (0.439–3.700) with a sensitivity interval still spanning 1. Two reviewers rated 53/62 and 50/62 assessable SQL checks met and 12/13 R checks met, with protocol and proxy disagreements remaining.
University of Missouri
Amir Habibdoust（First author, Corresponding author）
TrialCalibre1 Target Trial Emulation Benchmark, Expand, and Calibration (BenchExCal)Agentic System
arXiv Oct 8, 2026
CADFather coordinates learned, geometric, and optimization tools through a vision-language assistant, reaching zero invalid outputs across six CAD reconstruction benchmarks
CADFather is an autonomous, training-free agent in which a vision-language assistant decides which candidate CAD program to extend, which tool to invoke, and how many proposals to sample, unifying learned proposal generation, geometry-based algorithmic proposals, and parameter optimization in one CadQuery representation, and achieving zero invalidity with higher mean IoU and GMS and lower median Chamfer distance on the full DeepCAD, Fusion360, and MCB test sets.
CADFather is an autonomous, training-free agent in which a vision-language assistant decides which candidate CAD program to extend, which tool to invoke, and how many proposals to sample, unifying learned proposal generation, geometry-based algorithmic proposals, and parameter optimization in one CadQuery representation, and achieving zero invalidity with higher mean IoU and GMS and lower median Chamfer distance on the full DeepCAD, Fusion360, and MCB test sets.
Lomonosov Moscow State University Innopolis University
Gennadiy Savrasov（First author）
CADFather Mesh-to-CAD Reconstruction Agentic System reverse engineering
arXiv Oct 1, 2026
LiteReality-Agent recasts 3D reconstruction as code editing, reporting better geometry, realism, and simulation compatibility than Astra and Fable
LiteReality-Agent formulates reconstruction of interactable 3D indoor scenes from RGB-D scans as a coding problem: a coding agent gathers evidence with specialised tools and iteratively edits an executable Python script, within an observe-edit-verify harness covering evidence gathering, measurement, verification, layout optimisation, simulation readiness, and quality control, yielding digital twins reported as more geometrically accurate, visually realistic, and simulation-compatible than those from frontier models such as Astra and Fable.
LiteReality-Agent formulates reconstruction of interactable 3D indoor scenes from RGB-D scans as a coding problem: a coding agent gathers evidence with specialised tools and iteratively edits an executable Python script, within an observe-edit-verify harness covering evidence gathering, measurement, verification, layout optimisation, simulation readiness, and quality control, yielding digital twins reported as more geometrically accurate, visually realistic, and simulation-compatible than those from frontier models such as Astra and Fable.
Zhening Huang（First author）
LiteReality-Agent 3D Reconstruction Agentic System Digital Twin Coding agent
arXiv Jul 3, 2026
ShannonProver decomposes cryptographic main theorems into intermediate games and lemmas and generates EasyCrypt proof scripts, cutting expert proof work on ChaChaPoly1305 and MEE-CBC from weeks-to-months to hours
The work presents ShannonProver, an agentic system that, given a protocol and its security definitions modeled by a cryptographer, decomposes the main theorem into intermediate games and lemmas and constructs EasyCrypt proof scripts for those lemmas, evaluated on a new dataset of EasyCrypt lemmas spanning textbook primitives, deployed standardized protocols, and recent NIST proposals, completing on case studies such as ChaChaPoly1305 and MEE-CBC within hours proof developments that historically took experts weeks to months.
The work presents ShannonProver, an agentic system that, given a protocol and its security definitions modeled by a cryptographer, decomposes the main theorem into intermediate games and lemmas and constructs EasyCrypt proof scripts for those lemmas, evaluated on a new dataset of EasyCrypt lemmas spanning textbook primitives, deployed standardized protocols, and recent NIST proposals, completing on case studies such as ChaChaPoly1305 and MEE-CBC within hours proof developments that historically took experts weeks to months.
University of California, Berkeley University of Bristol
Yiping Ma（First author）
ShannonProver Agentic System EasyCrypt Automating Formal Cryptographic Proofs
arXiv Oct 1, 2026
PatchHolmes lets an agent read all 100 candidate commits at once, lifting patch-retrieval Recall@1 from 32.63% to 59.95%
PatchHolmes is a two-phase patch retrieval system whose first stage fuses BM25 with time decay and Qwen3-Embedding-8B dense retrieval via RRF into a top-100 candidate set, and whose second stage runs a frozen open-weight LLM agent that views all 100 candidates listwise through four tools (list_candidates, read_commit, read_file_diff, submit_answer) and submits a single best commit; on 809 CVEs from GitHubAD it reaches 59.95% Recall@1, 25.34% above the pointwise classifier Favia and 31.40% above IRCoT, adds 27.32% Recall@1 over the retriever's top pick on identical candidates, and transfers unchanged to PatchFinder_top10 to lift Recall@1 from 24.28% to 39.86%.
PatchHolmes is a two-phase patch retrieval system whose first stage fuses BM25 with time decay and Qwen3-Embedding-8B dense retrieval via RRF into a top-100 candidate set, and whose second stage runs a frozen open-weight LLM agent that views all 100 candidates listwise through four tools (list_candidates, read_commit, read_file_diff, submit_answer) and submits a single best commit; on 809 CVEs from GitHubAD it reaches 59.95% Recall@1, 25.34% above the pointwise classifier Favia and 31.40% above IRCoT, adds 27.32% Recall@1 over the retriever's top pick on identical candidates, and transfers unchanged to PatchFinder_top10 to lift Recall@1 from 24.28% to 39.86%.
Stevens Institute of Technology
Guanqun Yang（First author）
PatchHolmes Patch Retrieval Listwise Selection Agentic System Reciprocal Rank Fusion
arXiv Sep 25, 2026
OPGAgent uses a multi-tool agent with consensus to read panoramic dental X-rays, reaching 42.3% exact-match F1 on its OPG-Bench while holding false positives to 4.89 per case
The work proposes OPGAgent, a multi-tool dental agent planned by GPT-5.2 under the ReAct paradigm that orchestrates hierarchical evidence gathering, a specialized toolbox, and a consensus subagent, together with OPG-Bench, a structured-report protocol built on (Location, Field, Value) triples derived from real clinical reports; on OPG-Bench, comprising 1,009 anonymized OPGs and 5,219 VQA pairs, OPGAgent reaches 42.3% exact-match F1, a 49.7% aggregate score, 43.1% precision and 4.89 false positives per case, and leads MMOral-OPG at 62.53% accuracy.
The work proposes OPGAgent, a multi-tool dental agent planned by GPT-5.2 under the ReAct paradigm that orchestrates hierarchical evidence gathering, a specialized toolbox, and a consensus subagent, together with OPG-Bench, a structured-report protocol built on (Location, Field, Value) triples derived from real clinical reports; on OPG-Bench, comprising 1,009 anonymized OPGs and 5,219 VQA pairs, OPGAgent reaches 42.3% exact-match F1, a 49.7% aggregate score, 43.1% precision and 4.89 false positives per case, and leads MMOral-OPG at 62.53% accuracy.
National University of Singapore Monash University Monash Health University of Hong Kong Southeast University Melbourne Health
Zhaolin Yu（First author）Zongyuan Ge（Corresponding author）Litao Yang（Corresponding author）
OPGAgent OPG-Bench Dental Panoramic X-ray Interpretation Agentic System Consensus Subagent
arXiv Sep 17, 2026
Evidence-Grounded Agentic Formulation Development in an Autonomous Laboratory
The study presents Andromeda 2, an agentic system that reasons over structured in-house experimental evidence and invokes computational and experimental tools, achieving a 50% high-AUC hit rate for paclitaxel self-emulsifying drug delivery system (SEDDS) development at a matched budget of 96 formulations (versus 17% for Andromeda 1 and 2% for wet-lab DoE) and identifying 12 formulations meeting all four target product profile (TPP) objectives (versus 6 and 0), while an ablation showed that access to structured in-house evidence increased mean AUC by 34%.
The study presents Andromeda 2, an agentic system that reasons over structured in-house experimental evidence and invokes computational and experimental tools, achieving a 50% high-AUC hit rate for paclitaxel self-emulsifying drug delivery system (SEDDS) development at a matched budget of 96 formulations (versus 17% for Andromeda 1 and 2% for wet-lab DoE) and identifying 12 formulations meeting all four target product profile (TPP) objectives (versus 6 and 0), while an ablation showed that access to structured in-house evidence increased mean AUC by 34%.
Michael Craig（First author）
Andromeda 2 Agentic System Autonomous Laboratory Oral Paclitaxel Formulation

You’re all caught up