pith:HVZQWB6D
Additive Control Variates Dominate Self-Normalisation in Off-Policy Evaluation
An optimal additive baseline estimator asymptotically dominates self-normalized IPS in off-policy evaluation mean squared error.
arxiv:2602.14914 v3 · 2026-02-16 · cs.LG · cs.IR
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{HVZQWB6DJZM5KAUBJCRXRRPHMS}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
we prove that β∗-IPS, an estimator with an optimal additive baseline, asymptotically dominates SNIPS in Mean Squared Error. By analytically decomposing the variance gap, we show that SNIPS is asymptotically equivalent to using a specific -- but generally sub-optimal -- additive baseline.
The asymptotic dominance result assumes standard regularity conditions for OPE estimators (finite variances, bounded propensities, and existence of the optimal baseline) that are not detailed in the abstract and must hold for the variance decomposition to apply.
Optimal additive baseline IPS asymptotically dominates SNIPS in off-policy evaluation mean squared error.
Formal links
Receipt and verification
| First computed | 2026-07-28T01:22:29.926587Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
3d730b07c34e59d5028148a378c5e764ba77381e887d7517eceafa3c7e2dfaf3
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/HVZQWB6DJZM5KAUBJCRXRRPHMS \
| jq -c '.canonical_record' \
| python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: 3d730b07c34e59d5028148a378c5e764ba77381e887d7517eceafa3c7e2dfaf3
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "8b9737bb5bf00d562733b84c087ed81f149a3d75395ee16b6de5a6038d63bbfe",
"cross_cats_sorted": [
"cs.IR"
],
"license": "http://creativecommons.org/licenses/by/4.0/",
"primary_cat": "cs.LG",
"submitted_at": "2026-02-16T16:49:23Z",
"title_canon_sha256": "f75d5e9a94a14acb3dd4a1eb7e5729b76acd24c59369923ccd7022ff5328454e"
},
"schema_version": "1.0",
"source": {
"id": "2602.14914",
"kind": "arxiv",
"version": 3
}
}