File size: 3,023 Bytes
634e51e 6af6b10 634e51e 6af6b10 634e51e 6af6b10 634e51e 6af6b10 634e51e 6af6b10 634e51e 6af6b10 634e51e 6af6b10 634e51e 6af6b10 634e51e 6af6b10 634e51e 6af6b10 634e51e 6af6b10 634e51e 5f631f2 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 | {
"schema_version": 1,
"title": "Reproduction: Context-free Recognition with Transformers",
"emoji": "🧩",
"space_id": "snaykey/repro-gp-thompson-sampling-regret-bayesian-optimization",
"paper": {
"openreview_id": "JOyxs9ElI7"
},
"tags": [
"icml2026-repro",
"paper-JOyxs9ElI7"
],
"updated_at": "2026-07-31T00:00:00+00:00",
"root": {
"slug": "index",
"title": "Reproduction: Context-free Recognition with Transformers",
"file": "pages/index.md",
"children": [
{
"slug": "claim-1-theorem-3-1-general-cfl",
"title": "Theorem 3.1 proves that all context-free languages can be recognized by looped transformers using O(log(n)) looping layers and O(n^6) padding tokens (Section 3, Theorem 3.1).",
"file": "pages/claim-1-theorem-3-1-general-cfl/page.md",
"children": []
},
{
"slug": "claim-2-theorem-4-1-unambiguous",
"title": "Theorem 4.1 shows unambiguous context-free languages can be recognized with fewer resources than the general case: O(log^2(n)) looping layers and O(n^3) padding tokens (Section 4, Theorem 4.1).",
"file": "pages/claim-2-theorem-4-1-unambiguous/page.md",
"children": []
},
{
"slug": "claim-3-theorem-4-2-linear",
"title": "Theorem 4.2 further reduces requirements for unambiguous linear context-free languages to O(log(n)) looping layers and O(n^2) padding tokens by exploiting the linearity constraint (Section 4.2, Theorem 4.2).",
"file": "pages/claim-3-theorem-4-2-linear/page.md",
"children": []
},
{
"slug": "claim-4-lemma-4-1-pebbling",
"title": "Lemma 4.1 shows that Boolean formula evaluation can be computed in O(log(n)) steps via a parallel pebbling argument, which underlies the padding-token constructions used in the main theorems (Section 4, Lemma 4.1).",
"file": "pages/claim-4-lemma-4-1-pebbling/page.md",
"children": []
},
{
"slug": "claim-5-corollary-4-1-bfvp",
"title": "Corollary 4.1 shows that Boolean Formula Value Problem (BFVP) recognition requires zero padding tokens while retaining logarithmic depth (Section 4, Corollary 4.1).",
"file": "pages/claim-5-corollary-4-1-bfvp/page.md",
"children": []
},
{
"slug": "claim-6-table-1-tradeoff",
"title": "Table 1 summarizes the padding-versus-depth tradeoff across general, unambiguous, and linear-unambiguous context-free language classes, showing that grammatical restrictions systematically reduce required padding (Table 1).",
"file": "pages/claim-6-table-1-tradeoff/page.md",
"children": []
},
{
"slug": "executive-summary",
"title": "executive-summary",
"file": "pages/executive-summary/page.md",
"children": []
},
{
"slug": "conclusion",
"title": "conclusion",
"file": "pages/conclusion/page.md",
"children": []
}
]
}
} |