diwank/IBMDebaterArgQ
Viewer • Updated • 23.1k • 5
Diwank Tomer PRO
diwank
AI & ML interests
None yet
Recent Activity
liked a model 38 minutes ago
JetBrains/Mellum2-12B-A2.5B-Thinking reacted to sergiopaniego's post with 👍 39 minutes ago
most multi-turn RL loops have a silent bug: you decode the model's output to detect tool calls, then re-tokenize the conversation for the next turn. BPE isn't invertible, so decode then re-encode can land on different ids. gradient ends up on tokens the model never sampled. no crash, just quietly wrong math and broken training
@qgallouedec wrote a super educational blog on MITO (message-in, token-out) vs TITO (token-in, token-out) and how you might fix the problem above
go read it 🤓
https://qgallouedec-tito.hf.space/ liked a model about 6 hours ago
google/gemma-4-12B-itOrganizations
