moonshotai/Kimi-K3
Image-Text-to-Text • 2.8T • Updated • 2.85k • • 5.89k
Terra Nulles
TerraNull
·
AI & ML interests
None yet
Recent Activity
liked a model about 1 hour ago
moonshotai/Kimi-K3 upvoted a collection 29 days ago
DeepSpec liked a model about 1 month ago
deepseek-ai/DeepSeek-V4-Pro-DSparkOrganizations
None yet
liked a
model about 1 hour ago
upvoted a collection 29 days ago
liked a
model about 1 month ago
liked a
model 8 months ago
liked a
model 12 months ago
liked a
model over 1 year ago
upvoted a paper over 1 year ago
liked a
model over 1 year ago
liked a
model almost 2 years ago
liked a
model about 2 years ago
reacted to ehristoforu's post with 🤝 about 2 years ago
Post
2156
😐 Hello, there are a couple of interesting things. The first is that I will soon release several pretty cool SDXL models, the second is a little sad, I conducted long-term tests of training and merging of XL models and realized that XL will not improve soon, the architecture will not allow us to continue pushing realism and other interesting things into it, the entire community has brought XL closer to the maximum ideal on its architecture.
liked a
Space over 2 years ago
liked a
model over 2 years ago
Feature Request for Leaderboard: date added to hub
👍 4
8
#425 opened over 2 years ago
by 
madmaxbr5
liked 2
models over 2 years ago
liked a
Space over 2 years ago
reacted to vladbogo's post with 👍 over 2 years ago
Post
"A Closer Look at the Limitations of Instruction Tuning" is a new paper that explores the efficacy and limitations of Instruction Tuning (IT) in Large Language Models (LLMs) for conversational agents. The authors conduct a series of experiments using both LoRA fine-tuning (LFT) and standard full-parameter fine-tuning (SFT) across various LLMs and IT datasets.
The key findings are:
* LoRA fine-tuning (LFT) preserves the pre-training token distribution while SFT doesn't. This indicates that using LFT, post fine-tuning the model still heavily relies on the pre-training and doesn't acquire new information.
* Dataset scaling is ineffective for LFT - experiments show that scaling the dataset size 52x or even 326x doesn't improve the performance.
* LoRA fine-tuning mainly enhances response initiation and style without substantial knowledge enhancement.
* Full-parameter fine-tuning tends to degrade LLM knowledge base and increase hallucination occurrences.
* Popular other methods and adjustments fail to significantly outperform simple LoRA fine-tuned models in terms of conversational quality and accuracy.
Congrats to the authors @Sreyan88 and others for their work!
Paper: A Closer Look at the Limitations of Instruction Tuning (2402.05119)
The key findings are:
* LoRA fine-tuning (LFT) preserves the pre-training token distribution while SFT doesn't. This indicates that using LFT, post fine-tuning the model still heavily relies on the pre-training and doesn't acquire new information.
* Dataset scaling is ineffective for LFT - experiments show that scaling the dataset size 52x or even 326x doesn't improve the performance.
* LoRA fine-tuning mainly enhances response initiation and style without substantial knowledge enhancement.
* Full-parameter fine-tuning tends to degrade LLM knowledge base and increase hallucination occurrences.
* Popular other methods and adjustments fail to significantly outperform simple LoRA fine-tuned models in terms of conversational quality and accuracy.
Congrats to the authors @Sreyan88 and others for their work!
Paper: A Closer Look at the Limitations of Instruction Tuning (2402.05119)
- 2 replies
upvoted 2
papers over 2 years ago