6 2 15

hanzlajavaid PRO

hanzla

AI & ML interests

Direct Preference Optimization, Supervised Finetuning, Stable Diffusion

Recent Activity

posted an update about 10 hours ago

Reinforcement learning can sometimes lead to emergent behavior through much simpler training setups compared to large scale pre-training. I explored this idea by running a small GRPO experiment on Qwen3.5 4B, and the results were pretty exciting. Hypothesis: improving visual mathematical reasoning may also improve the model’s ability to transcribe LaTeX from images. I wrote a short breakdown of the experiment here: https://hanzlajavaid.github.io/blog/grpo-experiment-exploring-emergent-properties/

updated a model 8 days ago

hanzla/Qwen3.5-4B-mathvista-GRPO

published a model 8 days ago

hanzla/Qwen3.5-4B-mathvista-GRPO

View all activity

Organizations

Posts 5

Post

Reinforcement learning can sometimes lead to emergent behavior through much simpler training setups compared to large scale pre-training.

I explored this idea by running a small GRPO experiment on Qwen3.5 4B, and the results were pretty exciting.

Hypothesis: improving visual mathematical reasoning may also improve the model’s ability to transcribe LaTeX from images.

I wrote a short breakdown of the experiment here:
https://hanzlajavaid.github.io/blog/grpo-experiment-exploring-emergent-properties/

Post

2184

Hi community,

Few days back, I posted about my ongoing research on making reasoning mamba models and I found great insights from the community.

Today, I am announcing an update to the model weights. With newer checkpoints, the Falcon3 Mamba R1 model now outperforms very large transformer based LLMs (including Gemini) for Formal Logic questions of MMLU. It scores 60% on formal logic which is considered a tough subset of questions in MMLU.

I would highly appreciate your insights and suggestions on this new checkpoint.

Model Repo: hanzla/Falcon3-Mamba-R1-v0

Chat space: hanzla/Falcon3MambaReasoner