Issues

• [Optimization] Compile & profiling harness

  maida-ai/Research#50 · Filter by author z-a-f opened on Oct 15, 2025

• [Optimization] SSM cost knobs: proj_stride & optional low-rank B/C

  maida-ai/Research#49 · Filter by author z-a-f opened on Oct 15, 2025

• Pre-v0 Optimizations

  maida-ai/Research#33 · Filter by author z-a-f opened on Oct 14, 2025
  

• [Future] Small-World Attention (local cliques + random weak links)

  maida-ai/Research#20 · Filter by author z-a-f opened on Oct 2, 2025

• Release v0.1.0 — code, small checkpoints, and report

  maida-ai/Research#18 · Filter by author z-a-f opened on Oct 1, 2025 · OSS Polish, Leaderboards, and Docs

• CI: unit tests, formatting, and (optional) GPU smoke

  maida-ai/Research#17 · Filter by author z-a-f opened on Oct 1, 2025 · OSS Polish, Leaderboards, and Docs

• Example notebooks (train, streaming eval, profiling)

  maida-ai/Research#16 · Filter by author z-a-f opened on Oct 1, 2025 · OSS Polish, Leaderboards, and Docs

• Write method docs (DP-ASSM & BLADE) with diagrams

  maida-ai/Research#15 · Filter by author z-a-f opened on Oct 1, 2025 · OSS Polish, Leaderboards, and Docs

• Prepare Open LLM Leaderboard-style submission (or full model card)

  maida-ai/Research#14 · Filter by author z-a-f opened on Oct 1, 2025 · OSS Polish, Leaderboards, and Docs

• Publish a Hugging Face Space demo

  maida-ai/Research#13 · Filter by author z-a-f opened on Oct 1, 2025 · Evaluation Harness & Benchmarks

• System profiling: throughput, VRAM, and prefill speed

  maida-ai/Research#12 · Filter by author z-a-f opened on Oct 1, 2025 · Evaluation Harness & Benchmarks

• Long-document QA harness

  maida-ai/Research#11 · Filter by author z-a-f opened on Oct 1, 2025 · Evaluation Harness & Benchmarks

• Streaming curve harness (quality vs. length up to 1M tokens)

  maida-ai/Research#10 · Filter by author z-a-f opened on Oct 1, 2025 · Evaluation Harness & Benchmarks

• Infini-style mixed-path ablation

  maida-ai/Research#9 · Filter by author z-a-f opened on Oct 1, 2025 · Data, Training Scripts, and Baselines