Post-hoc auditing framework opens black box of neural auction mechanisms
CITransNet is unpacked using sparse autoencoders to verify features that align with Myerson's optimal auction theory.
Stories per day · 30 days
1 RSS
Relateddifferential-privacy1graph1llm-identity1mechanism-design1model-substitution1neural-auctions1summaries1watermarking1
15 stories, newest first · one card per story
Stories tagged #auditing
Researchers audit differentially private graph releases
New toolkit falsifies claimed privacy guarantees in Node-DP mechanism.
SUMMARK checks hosted LLM identity from reasoning summaries
A trusted auditor enrolls identities in advance; audit time uses only the claimed identity and the released summary, not provider logs.
CorrectGuard: Eyes-Off Correctness Estimation for Black-Box Security Guardrails
CorrectGuard estimates black-box AI guardrail correctness without inspecting user input or guardrail internals, boosting error-detection accuracy up to 25 points.
VStress: Correlation-Aware Auditing and Adaptive Budget Allocation for Repeated Verifiers
VStress-CA spends repeated verifier calls only when they add conditional information, improving accuracy under a fixed budget.
SKILLLITE: Evidence-Guided Malicious Skill Auditing with Compact LLMs
SKILLLITE audits third-party Agent Skills for hidden malicious behavior using compact local LLMs.
3
Selecting The Most Informative Tokens in Natural Language Autoencoders
Explaining five percent of tokens retains nearly all prompt-injection audit success on three of four datasets.
When Does Correction Become Repair? Mechanistic Auditing of Internal Interventions in Tool-Using LLMs
SAKIKO finds activation steering can raise tool-use scores while corrupting many previously correct LLM decisions.
FARE: Forensic Acceptance Region Estimation for Catching Bait-and-Switch Image Generators
FARE flags silent swaps of certified AI image generators by testing images against enrolled forensic artifacts.
Evidence-Grounded Auditing of Identification Assumptions in Climate-Policy Causal Evaluations
ARGUS, a language-model pipeline, audits identification assumptions in climate-policy difference-in-differences studies and abstains when evidence is missing.
Making Agents More Consistent: Skills Should Form Habits for Repeat Tasks
Paper proposes skill habit formation so agents cache deterministic skills from execution history, fixing 38-74% answer inconsistency and cutting tokens 14-56%.
Conformal Privacy Auditing: Calibrated Re-identification Attacks with Statistical Guarantees
Conformal Privacy Auditing certifies per-document re-identification risk against LLM attackers with statistical coverage guarantees.
AI labs want in-house auditors — but maybe they should shut the front door first
Security experts argue AI labs should prioritize agent sandboxing, monitoring, and network security basics over relying on third-party audits.
1
OPEN-1B: A Fully Auditable Training Run
Open-1B releases a 1B-parameter model with bitwise-reproducible training, letting independent auditors verify every step of the run on commodity hardware.
Scores Alone Do Not Prove Discovery: The Discovery Certification Protocol for Auditing AI Research Agents
Researchers introduce the Discovery Certification Protocol, an auditable test framework that verifies whether AI research agents' claimed discoveries are genuine.