NetGoodIndexSubmit a correction

Benefit Ledger · verified · Computer Science

AlphaDev sorting routines are added to LLVM libc++

DeepMind’s AlphaDev reinforcement-learning agent discovered faster short-sequence sorting algorithms that were merged into the LLVM libc++ sort library, published in Nature on 7 June 2023.

7 Jun 2023Tier 2 NotableMethodology 0.1

Current score

+0.96

3 base · Notable (tier 2 of 5, 3 pts)
× 0.9000 attribution · Primary causal contribution
× 0.7000 evidence · Peer review or independent validation
× 0.8500 realization · Realized outcome
× 0.6000 durability
Event-level product before credit split: 0.96

Notable deployed algorithm improvement (tier 2), not a new complexity class. High attribution. Peer review plus library merge. Realization is production libc++. Medium durability as compilers evolve.

What happened

AlphaDev treated assembly instruction selection as a game and found sort-3/4/5 routines that the authors reported as up to 70 percent faster for short sequences and about 1.7 percent faster for very long sorts in libc++. This was described as the first RL-discovered change to that part of the library. The gain is real but narrow.

Model attribution

+0.96

AlphaZero

RL agent that discovered assembly sorting routines later upstreamed.

The new routines originated from AlphaDev search.

Attribution 0.9000 · Credit share 100% · Google DeepMind

Claims

  • AlphaDev-discovered sort kernels were integrated into LLVM libc++.

    outcome · supported

Sources

primary sources

Secondary domains: Engineering

Revision history

  • 13 Sep 2026 · 0.00 → 0.96

    Initial adjudicated seed score under methodology 0.1.