Unverified50% confidenceBenchmarkExact time
A developer built a benchmark called Unlearning Security Audit in the Kaggle Benchmarking Challenge to test whether large language models can detect disguised unlearning code.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified安全研究已证明大语言模型能够自主执行渗透测试流程，包括信息收集、漏洞扫描、漏洞利用和权限提升68% similarUnverifiedThe probing classification method detects whether the hidden layers of a neural network encode specific linguistic or conceptual attributes by training simple classifiers.67% similarUnverifiedBenchmark hacking can artificially inflate benchmark scores through methods such as data contamination (test set leakage into training data), targeted training on the specific question formats of particular benchmarks, and selective reporting of the best benchmark results.66% similarUnverifiedCore tasks for large models in security offensive-defensive work include code auditing, vulnerability reproduction and exploit chain construction, and fuzzing assistance.65% similarUnverifiedAutomated vulnerability discovery typically relies on coverage-guided fuzzing tools such as syzkaller and AFL++, which have limited effectiveness against logic vulnerabilities.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/998821API
curl https://kongchang.com/api/v1/knowledge/claims/998821MCP
get_claim(id=998821)