nvidia/personaplex-7b-v1
Audio-to-Audio • Updated • 291k • 2.31k
Robert Vining
viningr
AI & ML interests
None yet
Recent Activity
liked a model 1 day ago
nvidia/personaplex-7b-v1 liked a model 11 days ago
Firworks/Hermes-4.3-36B-nvfp4 liked a model 20 days ago
microsoft/VibeVoice-ASR-HFOrganizations
None yet
liked a
model 1 day ago
liked a
model 11 days ago
liked a
model 20 days ago
liked a
model about 1 month ago
liked 2
models 2 months ago
liked 6
models 3 months ago
mradermacher/Huihui-Qwen3-VL-32B-Instruct-abliterated-i1-GGUF
33B • Updated • 2.23k • 19
huihui-ai/Huihui-Qwen3-VL-32B-Instruct-abliterated
Image-Text-to-Text • 33B • Updated • 1.34k • 27
Snowflake/snowflake-arctic-embed-l
Sentence Similarity • 0.3B • Updated • 48k • • 100
mradermacher/snowflake-arctic-embed-m-i1-GGUF
Feature Extraction • 0.1B • Updated • 68 • 1
mradermacher/Huihui-Qwen3-VL-30B-A3B-Thinking-abliterated-GGUF
31B • Updated • 615 • 2
mradermacher/Huihui-Qwen3-VL-30B-A3B-Thinking-abliterated-i1-GGUF
31B • Updated • 479 • 3
reactedto Teen-Different's post with 👀 3 months ago
Post
1718
Interesting... looked into Apple's DiffuCoder and the masked diffusion approach is actually hitting SOTA parity... basicallly proving global MDLM can work for code https://arxiv.org/pdf/2506.20639
but then you look at Tiny-A2D results and it’s the complete opposite...BD3LM (block diffusion) totally outperforms MDLM... and then both MDLM and BD3LM models struggle hard compared to the AR baselines... https://github.com/ZHZisZZ/dllm/tree/main/examples/a2d
digging into the why and i think it comes down to the adaptation method....tiny-A2D just SFT’d an AR model adaption to force it into diffusion... asking a model wired for left to right causal attention to suddenly think bidirectionally is a massive shock... it struggles to unlearn that strong AR inductive bias
...that explains why BD3LM worked better in their case... since it generates in chunks it preserves some sequential order... acts like a bridge or crutch that feels more natural to the original Qwen weights
contrast that with Apple... they didn't just SFT...they pre-trained/adapted on 130B tokens... fundamentally rewiring the model to understand global dependencies from the ground up
my theory is if we want MDLM to actually work we can’t just SFT... we need that heavy adaptation or full pre-training phase to break the causal priors... otherwise the model just gets confused
but then you look at Tiny-A2D results and it’s the complete opposite...BD3LM (block diffusion) totally outperforms MDLM... and then both MDLM and BD3LM models struggle hard compared to the AR baselines... https://github.com/ZHZisZZ/dllm/tree/main/examples/a2d
digging into the why and i think it comes down to the adaptation method....tiny-A2D just SFT’d an AR model adaption to force it into diffusion... asking a model wired for left to right causal attention to suddenly think bidirectionally is a massive shock... it struggles to unlearn that strong AR inductive bias
...that explains why BD3LM worked better in their case... since it generates in chunks it preserves some sequential order... acts like a bridge or crutch that feels more natural to the original Qwen weights
contrast that with Apple... they didn't just SFT...they pre-trained/adapted on 130B tokens... fundamentally rewiring the model to understand global dependencies from the ground up
my theory is if we want MDLM to actually work we can’t just SFT... we need that heavy adaptation or full pre-training phase to break the causal priors... otherwise the model just gets confused
liked a
model 3 months ago
liked 2
models 5 months ago
liked 2
models 6 months ago
upvoted an article 6 months ago
Article
20 New SDXL Fine Tuning Tests and Their Results (Better Workflow Obtained and Published)
•
3
liked a
model 6 months ago