Paper Detail

The Scaling Properties of Implicit Deductive Reasoning in Transformers

Enrico Vompa, Tanel Tammet

huggingface Score 6.4

Published 2026-05-05 · First seen 2026-05-09

General AI

Abstract

We 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.

Workflow Status

Review status
pending
Role
unreviewed
Read priority
later
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@misc{vompa2026scaling,
  title = {The Scaling Properties of Implicit Deductive Reasoning in Transformers},
  author = {Enrico Vompa and Tanel Tammet},
  year = {2026},
  abstract = {We 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.},
  url = {https://huggingface.co/papers/2605.04330},
  keywords = {Horn clauses, Transformers, implicit deductive reasoning, chain-of-thought, bidirectional prefix mask, algorithmic alignment, depth-bounded Transformers, huggingface daily},
  eprint = {2605.04330},
  archiveprefix = {arXiv},
}

Metadata

{}