Hugging Face's logo
John Smith's picture
🔄 In a Training Loop

John Smith PRO

John6666
569 2547 33197
MindOfJay's profile picture mkurman's profile picture danielfl's profile picture
·

AI & ML interests

None yet

Recent Activity

upvoted an article 44 minutes ago
Introducing the EdgeFirst Model Zoo
reacted to sebastientaylor's post with 🔥 44 minutes ago
The EdgeFirst Model Zoo is live — four YOLO families, detection and instance segmentation, measured on 20 real edge platforms. 837 validation sessions, with full Perfetto trace logs and COCO validation metrics. What should we measure next? Three polls, one click each: which platform to add, which model family, and which capability beyond detection. 👉 https://github.com/orgs/EdgeFirstAI/discussions/categories/polls 📖 https://huggingface.co/blog/EdgeFirst/model-zoo-intro 🤗 https://huggingface.co/spaces/EdgeFirst/Models
reacted to SeaWolf-AI's post with 👍 44 minutes ago
🔍 The attention mask stopped being an audit. An autoregressive model must not let position t depend on anything after t. Everyone checks this by inspecting the causal mask — but hybrid stacks now mix attention with state-space scans, and a scan has no mask. Every mask can be correct while information leaks through scans, aggregations, or normalization. ⚙️ So we test the property directly. Two inputs identical except at the last position, two forward passes, compare each layer's prefix, report the first layer that moves. No training, no gradients, no accelerator — seconds on CPU. 📊 Across 192 injected faults on eight checkpoints, mask inspection detected 0. The per-layer audit localized 192/192 to the exact layer. 🎯 Then we read the source before running anything. In transformers 5.7.0, the reference chunked scan reduces the inter-chunk recurrence over the input chunk axis; zamba2 and nemotron_h reduce over the output chunk axis. One axis. The dynamic audit confirmed the prediction exactly: Zamba2-1.2B leaks from length 256, its declared chunk size, and Nemotron-H-8B from 128, its declared chunk size. Bamba, Falcon-H1, Granite-4.0-H, Mamba2 and RecurrentGemma came back clean. ⚠️ Scope: the defect is on the PyTorch chunked-scan path, which runs whenever the fused kernels are absent — CPU, CI, stock installs. We could not build those kernels, so the fast path is untested and open. That caveat cuts both ways: a model can pass every fused-kernel test and still leak the moment it runs without them. 🧪 AX-RAY now carries this as its own axis. 39 models scored across causal, white-box and behavioral axes: 21 A, 3 B, 1 C, 14 F — with exactly 2 Causal-LEAK verdicts, the two the paper predicted. Badges separate a weights-level audit from an API-only one, so the two never get read as the same claim. 📄 https://arxiv.org/abs/2608.22876 🔬 https://huggingface.co/spaces/FINAL-Bench/AX-RAY 🤗 https://huggingface.co/papers/2608.22876
View all activity

Organizations

Glide's profile picture open/ acc's profile picture mekasiu's profile picture Solving Real World Problems's profile picture FashionStash Group meeting's profile picture No More Copyright's profile picture XORTRON - Criminal Computing's profile picture Future Tech's profile picture