Evidence · DiggingBeagle record
Four-billion-parameter decision model prioritizes memory-safety candidates with limited recall (research paper)
A small decision model assists deterministic vulnerability mining and agent review, but its poor patched-pair discrimination limits standalone use as a security verdict.
- Published
- Nov 12, 2026
- Source role
- primary disclosure
Evidence record
A small decision model assists deterministic vulnerability mining and agent review, but its poor patched-pair discrimination limits standalone use as a security verdict.
Claim-level citations (4)
- supportsFour-billion-parameter decision model prioritizes memory-safety candidates with limited recall: For a held-out set of 75 known CVEs, the authors report that 72 were represented in the input corpus, only 29 were marked by the preliminary miner and 18 survived subsequent candidate limits; 16 reached the final top-25 list. Most misses occurred before the small decision model could rank the evidence.
Section 6.3, known-CVE discovery pipeline, candidate-mining and top-25 recall breakdown
- supportsFour-billion-parameter decision model prioritizes memory-safety candidates with limited recall: Authors describe 187 internally reviewed memory-safety findings, yet independent-project known-CVE top-25 ranking recovers only 16 of 75 test defects and smart-contract adaptation yields 61 false positives with zero confirmed findings.
Evaluation, campaign counts, known-CVE ranking and EVM adaptation
- supportsFour-billion-parameter decision model prioritizes memory-safety candidates with limited recall: Deterministic candidate miners and suspicious-write locators shortlist code fragments; Intern-Decision-4B then ranks vulnerability and reachability questions before larger coding agents and human reviewers investigate.
Methods, deterministic miners, localizers and decision-model scoring
- supportsFour-billion-parameter decision model prioritizes memory-safety candidates with limited recall: The decision model scores native C/C++ vulnerable-versus-patched pairs with ROC-AUC 0.564, and at a threshold of 0.5 detects only 15 percent of vulnerable examples with 6 percent benign false positives; that is inadequate as a standalone vulnerability verdict.
Section 6, Table 2 native C/C++ vulnerability and patched-pair evaluation