{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreibxaj4hzf5je6kutz2a2msivgonop63g32gfxl5sbfut3yvnnhcrm",
"uri": "at://did:plc:3fychdutjjusoqeq24ljch6q/app.bsky.feed.post/3mlcogwh2sfa2"
},
"coverImage": {
"$type": "blob",
"ref": {
"$link": "bafkreiflo6xt7is6b2iafwghkjahlgggocme5jwjsbeuqqwcywuvjhmszm"
},
"mimeType": "image/png",
"size": 24783
},
"path": "/abs/2605.04330v1",
"publishedAt": "2026-05-07T00:00:00.000Z",
"site": "https://arxiv.org",
"tags": [
"Enrico Vompa",
"Tanel Tammet"
],
"textContent": "**Authors:** Enrico Vompa, Tanel Tammet\n\nWe investigate the scaling properties of implicit deductive reasoning over Horn clauses in depth-bounded Transformers. By systematically decorrelating provability from spurious features and enforcing algorithmic alignment, we find that in sufficiently deep models with a bidirectional prefix mask, implicit reasoning approaches explicit CoT performance across graph topologies and problem widths, though CoT remains necessary for depth extrapolation.",
"title": "The Scaling Properties of Implicit Deductive Reasoning in Transformers"
}