bkhmsi/micro-llama-3b
12B • Updated • 5 • 2
Mike Ravkine PRO
mike-ravkine
AI & ML interests
LLM Research / Development / Evaluation
Recent Activity
liked a model 1 day ago
bkhmsi/micro-llama-3b liked a model 3 days ago
deepreinforce-ai/Ornith-1.0-397B-FP8 liked a model 21 days ago
ideogram-ai/ideogram-4-fp8Organizations
liked a
model 1 day ago
liked a
model 3 days ago
liked 3
models 21 days ago
liked 2
models about 1 month ago
liked 2
models about 2 months ago
upvoted a paper about 2 months ago
liked 5
models 2 months ago
ibm-granite/granite-4.1-8b
9B • Updated • 293k • 202
ibm-granite/granite-4.1-30b
Text Generation • 29B • Updated • 73k • 134
poolside/Laguna-XS.2
Text Generation • 33B • Updated • 84.4k • 317
deepseek-ai/DeepSeek-V4-Flash
Text Generation • 158B • Updated • 1.93M • • 1.64k
SandyResearch/parcae-1.3b
Text Generation • Updated • 985 • 7
replied to their post 2 months ago
Some clarity is emerging:
The distribution of response lengths has shifted considerably in 3.6 and 2 of my tasks are no longer fitting into 16k, the ignorance zone blows up.
Re-running at 32k then we'll see if that extra thinking pays off or nah.
An interesting outlier here is the word-sort task where 3.6 thinks ~half as much and this costs it about 10pp of performance.
posted an update 2 months ago
Post
110
A word of warning that my initial run of 3.6 produced unfavorable results vs 3.5; performance on out-of-distribution and instruction following tasks appears to have collapsed. Potentially a vLLM 0.19 issue here, the original eval was done with their fork of 0.18. I am re-running both with nightly so we have apples to apples and will report back, curious to get to the bottom of this one.
- 1 reply
liked 3
models 3 months ago
