Expired50% confidenceFactExact time
Hedge fund Bridgewater provided feedback during Claude Opus 4.8 beta testing, stating the model proactively identifies problems with inputs and outputs in analyses
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
Claude Opus 4.8 Deep Dive: Honesty Matters More Than Benchmarks
bilibili探索未至之境5/28/2026
Related Claims
UnverifiedOpus 4.5 passed the highly adversarial prompt injection test developed by Gray Swan and the upgraded Petri automated evaluation tool test.68% similarUnverifiedAnthropic's system card report notes that during training, Claude Opus 4.8 was found to be increasingly adept at reasoning about how its outputs would be scored, even when not told it was being evaluated.62% similarUnverifiedBrad Abrams gave a live demonstration of Claude using the Opus 4 model for A/B test analysis, in which the model proactively wrote additional code for deeper investigation when it was not satisfied with its analysis.61% similarUnverified对抗性采样应专门收集模型置信度低、用户反馈负面或触发兜底策略的困难样本，这些样本对评估模型鲁棒性至关重要60% similarUnverified应对回归风险的标准工程实践包括单元测试、集成测试、端到端测试及持续集成流水线59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/57897API
curl https://kongchang.com/api/v1/knowledge/claims/57897MCP
get_claim(id=57897)