Unverified50% confidenceFactExact time
The paper points out four evaluation hazards that arise when comparing the two versions.
1
Sources
50%
Confidence
Long-term
Relevance
11/6/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedCommonly used metrics for assessing tail risk include Conditional Value at Risk (CVaR), worst-case analysis, and adversarial testing.70% similarUnverifiedThe study employs four evaluation metrics: accuracy (Acc), hindsight trap rate (HTR), answer instability rate (AIR), and hindsight bias rate (HBR).68% similarUnverifiedThe researchers propose a risk-sensitive evaluation framework that assesses the patient safety risks of medical large models through six categories of risk language (treatment instructions, contraindications, dosage expressions, emergency triage prompts, high-alert medication mentions, overconfident assertions) and their severity scores.66% similarUnverifiedThe study recommends that benchmarks should be run multiple times and report variance intervals, explicitly marking which comparative conclusions are statistically inconclusive64% similarUnverifiedThe researchers selected four candidate datasets from HELM Safety that may target "harmful refusal", three of which are already saturated and cannot provide effective discriminating information.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1142300API
curl https://kongchang.com/api/v1/knowledge/claims/1142300MCP
get_claim(id=1142300)