Extension: Om Nepal, Sushant Aryal, Oluseyi Olukola, Nick Rahimi, Metrics Failure in LLM-Based Code Vulnerability Repair: An Empirical Study and a Change-Aware Screen (arXiv:2609.26749, 2026)

Record: kaal:position:2026-09-23-001 · 2026-09-23

Kaal's account of threshold gaming extends to automated evaluation: controlled vulnerability-repair experiments show that optimizing an administrable success proxy can raise measured performance while reducing alignment with the underlying objective and producing outputs that satisfy the proxy without accomplishing the repair.

• where a proxy metric becomes an optimization target
• where satisfying the measured threshold can be separated from achieving the substantive objective
• limited to the source's tested vulnerability-repair models, prompts, dataset, and evaluation harness
• The position is Kaal's own cross-domain application of the cited claim; the source's authors do not make it.
• The source passage is taken from the source abstract; the full paper was not reviewed.
• The source is an empirical study of LLM-based C/C++ single-function vulnerability repair, not funds or securities regulation.

Om Nepal, Sushant Aryal, Oluseyi Olukola, Nick Rahimi, Metrics Failure in LLM-Based Code Vulnerability Repair: An Empirical Study and a Change-Aware Screen (arXiv:2609.26749, 2026)

kaal:claim:1428387-036
Kaal, Hedge Fund Valuation Retailization, Regulation, and Investor Suitability (2009). SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1428387
Source PDF sha256: 6aa3a280dc6750723be2389f3af2aabf3a16c4ce20e49fcdaddd7f2ea95a67aa

risk-and-incentivesregulatory-failureai-and-agentsempirical-evidence

Affirmed in kaalvis:2026-09-23:connection-discovery-two-v1 on 2026-09-23. Review record.

Canonical markdown sha256: 92e2bdb2c4a51a1cd81a09044d6c7b78352a5d203d65f7349a12973c1b81e459
curl -s https://wulfkaal.github.io/positions/2026-09-23-001.md | sha256sum