A small decision model assists deterministic vulnerability mining and agent review, but its poor patched-pair discrimination limits standalone use as a security verdict. Deterministic candidate miners and suspicious-write locators shortlist code fragments; Intern-Decision-4B then ranks vulnerability and reachability questions before larger coding agents…
Inspect the ClaimsCase · DiggingBeagle record
Four-billion-parameter decision model prioritizes memory-safety candidates with limited recall
A small decision model assists deterministic vulnerability mining and agent review, but its poor patched-pair discrimination limits standalone use as a security verdict.
Research scan of 4,245 selected code repositories with internal review and ASan-style reproduction; undisclosed candidate defects remain under coordination.
Assessment profile
Assessment method- Setting
- evaluation
- Exploitation
- not established
- Evidence
- Grade C
- Remediation
- not applicable
- AI role
- WITH AI
Basis and provenance
The campaign reports reproduced findings, but independent known-CVE ranking is limited and the EVM adaptation found no confirmed defects; exploitability and production impact remain unverified. Classified as evaluation with exploit status not_established. Evidence grade C reflects the original paper or maintainer advisory and its stated limitations, not an independent audit of all deployments. Comparable realized harm is not established, so impact stays unrated; remediation status not_applicable is limited to the cited experiment or affected software version. New source-specific boundary: Increasing the model's classification accuracy alone cannot recover a defect discarded during initial repository mining or candidate-budget selection.
Assessed Nov 15, 2026 using diggingbeagle.assessment/1.
Evidence basis: CLM-R245-SMALL-MODEL-VULN-PRIORITY-01 · CLM-R245-SMALL-MODEL-VULN-PRIORITY-02 · CLM-R246-SMALL-MODEL-VULN-PRIORITY-BOUNDARY · CLM-ENR250-16 · Four-billion-parameter decision model prioritizes memory-safety candidates with limited recall (research paper)
At a glance
Reproduced crashes do not establish exploitability on an installed entry point, and the withheld findings cannot yet be independently counted or verified. Model rankings are not automatic fixes. No evidence here establishes the number of production victims or a broader failure rate outside the cited evaluation.
Limits and uncertaintyFull account
A small decision model assists deterministic vulnerability mining and agent review, but its poor patched-pair discrimination limits standalone use as a security verdict.
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. 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.
Reproduced crashes do not establish exploitability on an installed entry point, and the withheld findings cannot yet be independently counted or verified. Model rankings are not automatic fixes.
The harness extracts incomplete-fix candidates and suspicious writes with deterministic miners, then asks a four-billion-parameter model to rank where expensive code-agent review should go. The model does not establish exploitability independently, and the reported 187 internally reviewed memory-safety findings cannot yet be audited externally because details remain under coordinated disclosure. A prioritized queue is useful only insofar as the reviewers understand what the ranking misses.
In held-out known-CVE testing, the small model separates vulnerable code from benign code better than it separates a vulnerable implementation from its own repaired variant. The 0.564 patched-pair AUC and 15 percent recall at the cited threshold prevent a claim that it can certify safety. Its attempt to transfer the same method to smart contracts produced 61 false positives and zero confirmed findings, consistent with the need for cross-function business-logic reasoning. The contribution is a cheap triage stage, not autonomous vulnerability discovery or general EVM security assurance.
Mechanism and trust boundary
- 01
Entry or starting condition
Deterministic miners and localizers shortlist suspicious source locations across selected code repositories.
Boundary: attacker or research precondition
- 02
Crossing the relevant boundary
Intern-Decision-4B ranks candidate vulnerabilities and reachability questions before stronger tools and human reviewers investigate.
Boundary: mechanism and authorization boundary
- 03
Observed or reported result
The campaign reports reproduced findings, but independent known-CVE ranking is limited and the EVM adaptation found no confirmed defects; exploitability and production impact remain unverified.
Boundary: measured outcome and evidentiary limit
Claims & evidence
CLM-ENR250-16For 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.supported
Basis: reported finding
- supportsFour-billion-parameter decision model prioritizes memory-safety candidates with limited recall (research paper)primary disclosure
Section 6.3, known-CVE discovery pipeline, candidate-mining and top-25 recall breakdown
CLM-R245-SMALL-MODEL-VULN-PRIORITY-01Authors 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.supported
Basis: reported finding
- supportsFour-billion-parameter decision model prioritizes memory-safety candidates with limited recall (research paper)primary disclosure
Evaluation, campaign counts, known-CVE ranking and EVM adaptation
CLM-R245-SMALL-MODEL-VULN-PRIORITY-02Deterministic 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.supported
Basis: reported finding
- supportsFour-billion-parameter decision model prioritizes memory-safety candidates with limited recall (research paper)primary disclosure
Methods, deterministic miners, localizers and decision-model scoring
CLM-R246-SMALL-MODEL-VULN-PRIORITY-BOUNDARYThe 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.supported
Basis: reported finding
- supportsFour-billion-parameter decision model prioritizes memory-safety candidates with limited recall (research paper)primary disclosure
Section 6, Table 2 native C/C++ vulnerability and patched-pair evaluation
Unknowns and contradictions
- Reproduced crashes do not establish exploitability on an installed entry point, and the withheld findings cannot yet be independently counted or verified. Model rankings are not automatic fixes.
- No evidence here establishes the number of production victims or a broader failure rate outside the cited evaluation.
Sources and citation
Material revision history
- Nov 15, 2026 · Canonical change recorded · new in release · revision 251