👋 Open to Work
Andrew Thompson
AndrewThompson1233
AI & ML interests
Deep Learning / Model Architecture & Training
Recent Activity
new activity 24 days ago
AndrewThompson1233/maba-v1-architecture:Factorized embeddings intervention for GPT-2 new activity 26 days ago
AndrewThompson1233/maba-instant-v1:Model? new activity 26 days ago
AwareLiquid/M2-2B:Horizon bottleneck at 128-byte sequence length and associative state updatesOrganizations
Factorized embeddings intervention for GPT-2
17
#1 opened about 1 month ago
by 
gpjt
Model?
3
#1 opened 28 days ago
by 
ArushCodes
Horizon bottleneck at 128-byte sequence length and associative state updates
5
#1 opened about 2 months ago
by AndrewThompson1233
Mitigating chat overfitting on small corpora via block recycling and RoPE
14
#1 opened about 2 months ago
by AndrewThompson1233
Expert orthogonalization horizons, depth-upscaling continuity, and context memory walls on 8GB VRAM
3
#1 opened about 1 month ago
by AndrewThompson1233
State attenuation across depth and long-context needle retention in 3:1 NoPE hybrids
4
#1 opened about 1 month ago
by AndrewThompson1233
Overcoming the ARC-Easy floor and vocabulary budget on an 80M footprint
❤️ 1
8
#1 opened about 2 months ago
by AndrewThompson1233
The 256-token context cliff and sequence scaling in custom 346M chat architectures
10
#1 opened about 1 month ago
by AndrewThompson1233
Numerical validation rigor, the 17.8% vocabulary tax, and discourse-level perplexity cliffs in 46M SLMs
❤️ 1
2
#1 opened about 1 month ago
by AndrewThompson1233
The 39% vocabulary parameter tax and layer allocation on Colab T4 budgets
3
#1 opened about 1 month ago
by AndrewThompson1233
Bilinear pointer routing, schema scaling in MCU memory budgets, and coordinate indexing in sub-1M AFMs
#1 opened about 1 month ago
by AndrewThompson1233
The 18.6% vocabulary tax, vertical shortcut saturation, and multi-turn intent drift in 125M architectures
#1 opened about 1 month ago
by AndrewThompson1233
Runaway generation mechanics, rotary phase mismatch, and state dissipation in 3:1 KDA-MLA hybrids
2
#1 opened about 1 month ago
by AndrewThompson1233
Value residual scaling, the 25.8% vocabulary tax, and horizontal state grounding in deep-thin SLMs
2
#1 opened about 1 month ago
by AndrewThompson1233
Memory bandwidth saturation, KV-cache thermal throttling, and O(1) sequence scaling on edge devices
2
#1 opened about 1 month ago
by AndrewThompson1233
Continuous associative fast-weights and the 512-token context horizon in lifelong learning
2
#1 opened about 1 month ago
by AndrewThompson1233
The 67M untied vocabulary footprint and context scaling on 6GB RTX 3050 budgets
2
#1 opened about 1 month ago
by AndrewThompson1233
Subspace entanglement in attention o_proj edits and downstream expert routing drift
2
#1 opened about 1 month ago
by AndrewThompson1233