Anthropic model refutes 3SUM and APSP hypotheses with subquadratic algorithms
Anthropic’s internal research model has produced the first deterministic algorithms that break the long‑standing quadratic and cubic barriers for two classic problems. The new method solves 3SUM on n integers in O(n^{1.9992}) time and all‑pairs shortest paths on directed n‑vertex graphs with polynomially bounded integer weights in O(n^{2.9995}) time. By exploiting a thin‑matrix product…
Key points
- Anthropic’s internal model solves 3SUM in O(n^{1.9992}) time, beating the quadratic barrier
- The same technique yields APSP on directed graphs in O(n^{2.9995}) time, refuting the APSP hypothesis
- The approach uses a thin‑matrix product algorithm with D ≤ N^{1/18}, achieving O(N^2/D^{0.063}) operations for selected entries
The authors describe the algorithm as a modification of Coppersmith’s rectangular matrix multiplication, using a ten‑multiplication identity of Schönhage to limit work to only the needed entries. They also present a data‑structure version that can answer single‑entry queries on the fly. The results are posted as a pre‑print on arXiv, with accompanying code and data for community verification.
Internal Anthropic model refuted the 3SUM and APSP hypotheses: first truly subquadratic 3SUM and truly subcubic APSP (Alman & Vassilevska Williams)
arxiv.org · 6 October 2026
Loading the full article…
This text was published by arxiv.org. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
1source- Reddit discussionreddit.com
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Opinion: Author explains difference between Strong AI and Weak AI · 1 src
- PhAI Labs introduces JEPA-Anything framework for cross-domain world models · 1 src
- Llama.cpp adds MTP decoding for Qwen4Exp and GLM-5.3-Flash hybrid model · 2 src
- GitHub releases ReviewBench to evaluate AI code reviews across 219 pull requests · 2 src
- Researchers test fixed token codes for language models at 100B-token scale · 1 src
Comments
via GitHub Discussions