๊ฐ€์ž… ํ›„ ์ดˆ๋Œ€ ๋งํฌ๋ฅผ ๊ณต์œ ํ•˜๋ฉด ๋™์˜์ƒ ์žฌ์ƒ ๋ฐ ์ดˆ๋Œ€ ๋ณด์ƒ์„ ๋ฐ›์„ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

cv usk
@cv_usk
AI / Software Research Notes AI Agent, LLMOps, MLOps, Software Architecture ๆŠ•็จฟใฏๅ€‹ไบบใฎๆ„่ฆ‹ใงใ™ใ€‚
๊ฐ€์ž… May 2026
286 ํŒ”๋กœ์ž‰ ์ค‘    474 ํŒฌ
TL;DR: Causal foundation models can report sound intervals per unit, yet become overconfident on group averages. The cause is aggregation that ignores correlated unit errors. Unit-level aggregation breaks coverage in causal foundation models Key points ใƒป๐ŸŽฏ For the full-sample average effect on ACIC 2016, CausalPFN's calibrated 95% interval contained the truth in only 8 of 40 datasets ใƒป๐Ÿงฎ The cause is correlated unit errors: variance is inflated by 1+(kโˆ’1)ฯ (the design effect), yet units were summed as if independent ใƒป๐Ÿ” The fix is SCM-marginalized aggregation: units within a draw share one causal model ใƒป๐Ÿง  Approximated with small CNP and latent NP models, also usable as heads on a frozen CausalPFN ใƒป๐Ÿ“ˆ Full-sample coverage across 240 datasets on 6 benchmarks is about 90% (CausalPFN: 53%) ใƒปโš ๏ธ The cost: lower unit-level accuracy on 3 benchmarks and wider intervals If decisions rely on group averages, how uncertainty is aggregated deserves as much scrutiny as the point estimate. #CausalInference# #FoundationModels#
๋” ๋ณด๊ธฐ