DeltaNet-PC: predictive coding offers native test-time memorization for sequences
fkasummer · x · 2026-10-20
DeltaNet-PC is out, showing that predictive coding natively supports test-time memorization of sequential data. The key insight: treat updates to fast associative memory as an inferred latent variable rather than a long-term learned weight — a potentially new architectural angle on rapid in-context memory.
More from Research
- UC Berkeley Study: Coding Harnesses Barely Move Success Rates but Swing Costs 5x — trashacct383 · 2026-10-20
- Technical Handbook Breaks Down Decision Models for AI Agents via Drex 2.0 — techNmak · 2026-10-20
- Charlie Marsh: models can mine published literature for algorithms, one-shotting a 12,000x speedup — Vjeux · 2026-10-20
- Position paper: humans are missing from AI coding agent research — math_rachel · 2026-10-20
- Hazy Research's open-source NCM generates text 31x faster via coarse-to-fine concept modeling — simran_s_arora · 2026-10-20
- New scaling-law paper argues data curation cannot be compute-agnostic: repeated high-quality data loses utility fast — cwolferesearch · 2026-10-20