Safe Portfolio Learning paper: untrusted ML predictions match Safe benchmark within 15.9 bps, no alpha
PtrPomorski · x · 2026-11-07
A paper shared by PtrPomorski, "Safe Portfolio Learning from Untrusted Predictions," proposes a controller that chooses among Safe, Follow, and Reverse portfolio actions when prediction direction, reliability, and timing are all untrusted. It combines a share-adaptive signed master, an abstention-capable evidence gate, a delayed collateral budget, and a protected split account, with dual certificates: regret tracking under delay and one-sided intervention, plus a pathwise collateral safety bound that holds when predictions fail. Empirically it's conservative: on a 24-stock development universe the overlay nears equal-weight performance at a fraction of the turnover; on a frozen 82-stock holdout the pre-specified Reverse orientation fails to beat equal weight, and flipping the sign to Follow doesn't recover an edge either. Protected wealth stays within 15.9 basis points of the Safe benchmark throughout. The contribution is safe adaptive use of untrusted predictions, not demonstrated alpha.
More from Research
- MCPHallu: 14 frontier MCP agents average 0.642, only 23% admit infeasible goals — dair_ai · 2026-11-07
- Nature Cancer's DeepFAN: AI nodule reading aids judged by how they change radiologists — jiqizhixin · 2026-11-07
- Once AI can simulate a human cell, aging becomes a software problem — rand_longevity · 2026-11-07
- Is generative drift a fundamental flaw in AI? Reddit debates agent reliability — Peacefulhuman1009 · 2026-11-07
- ICLR authors: rebut nitpicking reviewers by quoting their exact claims — brunofmr · 2026-11-07
- CIKM 2026 tutorial on Agentic Search covers RL optimization, deep research, and evaluation — mrdrozdov · 2026-11-07