An NLA trained with random-length truncation of the verbalizer's output, so the most important information comes first. Checkpoints, data, baseline.
Smitty PRO
syvb
·
AI & ML interests
None yet
Recent Activity
updated a model about 3 hours ago
syvb/nla-qwen2.5-7b-L20-suffix-rl published a model about 3 hours ago
syvb/nla-qwen2.5-7b-L20-suffix-rl updated a dataset about 6 hours ago
syvb/nla-weirdchat-explanationsOrganizations
None yet
An NLA trained with random-length truncation of the verbalizer's output, so the most important information comes first. Checkpoints, data, baseline.
16 repeated injection markers w/ per-slot learned affine vs 1-slot control. Warm-started from nanonla-qwen3-8b-L24-av.
v2 item-truncation matryoshka NLA (Qwen2.5-7B L20): warm-start AV verbalizer + AR critic checkpoints, warm-start data, base NLA.