PRO-LONG
Programmatic memory for long-horizon LLM agents: the harness appends everything to one log, and the agent searches it with code. 97.4% on ARC-AGI-3 (arXiv:2607.20064)
- 465
- Python
- MIT
- 3 wk ago
Programmatic memory for long-horizon LLM agents: the harness appends everything to one log, and the agent searches it with code. 97.4% on ARC-AGI-3 (arXiv:2607.20064)
A Mimetic Procedural Benchmark Generator for the Abstraction and Reasoning Corpus
ARC-AGI-2 solver using geometric embeddings, hyperbolic rule inference, and adversarial structure probing
Deterministic no-float Math Court MCP clients for agent builders. Bind affine.earth. Integers sealed. Floats refused. Server is not in this git.
Code for "The role of positional encodings in the ARC benchmark" (arXiv:2502.00174)
Wild ideas, experiments, and explorations for solving the ARC-AGI Challenge