Lucas Atkins PRO
Crystalcareai
AI & ML interests
MoE Models & Dataset Curation
Recent Activity
updated
a model
8 days ago
arcee-ai/Virtuoso-Lite
reacted
to
sometimesanotion's
post
with 👍
9 days ago
**Update** Either I had some wrong numbers plugged in to estimate benchmark numbers from comparator, or the benchmark changed. Virtuoso Small v2 at 41.07 average is still very impressive, especially for writing draft copy for business purposes, while Lamarck remains a chatty generalist-reasoning model.
I've felt confident that 14B Qwen finetunes and merges could break the 42.0 average, and Arcee **came close** with https://huggingface.co/arcee-ai/Virtuoso-Small-2. Congratulations to @arcee-ai!
Just two months ago, it was easy to think that 14B had plateaued, that you could have high IFEVAL or high MUSR/MATH/GPQA at 14B, but not both. That barrier is completely shattered. I see a pathway to even better, and Virtuoso Small 2 is a big part of why. Very impressive work. This community would expect no less from Arcee.
Just look at this graph! Keep in mind, my merges here build on the first Virtuoso Small, and *-DS merges build on DeepSeek R1. There are some impressive merges in the pipe!
reacted
to
sometimesanotion's
post
with 🚀
9 days ago
**Update** Either I had some wrong numbers plugged in to estimate benchmark numbers from comparator, or the benchmark changed. Virtuoso Small v2 at 41.07 average is still very impressive, especially for writing draft copy for business purposes, while Lamarck remains a chatty generalist-reasoning model.
I've felt confident that 14B Qwen finetunes and merges could break the 42.0 average, and Arcee **came close** with https://huggingface.co/arcee-ai/Virtuoso-Small-2. Congratulations to @arcee-ai!
Just two months ago, it was easy to think that 14B had plateaued, that you could have high IFEVAL or high MUSR/MATH/GPQA at 14B, but not both. That barrier is completely shattered. I see a pathway to even better, and Virtuoso Small 2 is a big part of why. Very impressive work. This community would expect no less from Arcee.
Just look at this graph! Keep in mind, my merges here build on the first Virtuoso Small, and *-DS merges build on DeepSeek R1. There are some impressive merges in the pipe!
Organizations
Crystalcareai's activity
Add Apache-2.0 license :3
#1 opened 13 days ago
by
SaisExperiments
This cross-architecture distillation, with Phi?
2
#14 opened 26 days ago
by
sometimesanotion
Context size
1
#6 opened 2 months ago
by
LoSboccacc
Chat Template
2
#5 opened 2 months ago
by
isr431
Question about model's origin
2
#2 opened 2 months ago
by
sometimesanotion
Fix tokenizer.json with file from Qwen/Qwen2.5-14B
1
#3 opened 2 months ago
by
MariusNocturnum
use the original Qwen2.5-14B-Instruct tokenizer
1
#4 opened 2 months ago
by
MaziyarPanahi
Adding Evaluation Results
#1 opened 2 months ago
by
leaderboard-pr-bot
add Instruct datasets and base model
#1 opened 2 months ago
by
Crystalcareai
Chạy trên nhiều GPU cùng 1 lúc để cải thiện hiệu năng
1
#1 opened 2 months ago
by
sieudd
Update SuperNova-Medius with a merge with Qwen/Qwen2.5-Coder-14B-Instruct + Further Training 😋
11
#12 opened 3 months ago
by
Joseph717171
max output tokens?
1
#11 opened 3 months ago
by
sirus
Update _name to "arcee-ai/SuperNova-Medius"
#5 opened 4 months ago
by
ggbetz
Adding Evaluation Results
#3 opened 4 months ago
by
leaderboard-pr-bot
llama.cpp convert problem report(about `tokenizer.json`)
6
#2 opened 4 months ago
by
DataSoul
2 base models = a nice merge UI on the model page
2
#1 opened 4 months ago
by
victor
Explain these Benchmark Results
2
#2 opened 4 months ago
by
Joseph717171
Distill Llama-3.2-1B-Instruct from Llama-405B-Instruct to make SuperNova-Pico
1
#14 opened 5 months ago
by
Joseph717171
Adding Evaluation Results
#15 opened 4 months ago
by
CombinHorizon
Why is the tokenizer.json not the same as LLaMa-3.1-8B-Instruct
1
#6 opened 5 months ago
by
Joseph717171