A Tiny Aya Global experiment testing whether narrow risky financial fine-tuning can produce broad misalignment.
Mrinaal Arora
mrinaalarora
AI & ML interests
AI alignment, AI safety, multilingual models and agent evaluation.
Recent Activity
updated a collection 22 days ago
Tiny Aya Emergent Misalignment updated a collection 22 days ago
Tiny Aya Emergent Misalignment updated a collection 22 days ago
Tiny Aya Emergent MisalignmentOrganizations
None yet
spaces 6
pinned
Running
Agents
Trackio
🎯
Show training metrics for your LLM project
Sleeping
RL
DryLabSim
🚀
LLM agents plan noisy biological experiment pipelines
Running
RL
Crisisops Environment Server
🎛
Manage crisis response units in a dynamic simulation
Sleeping
RL
Textarena Environment Server
👀
Interacting with TextArena games via a web chat interface
Sleeping
RL
Json Cleaning Env Environment Server
🖥
Clean messy JSON to fit a target schema
Sleeping
Agents
2
Qwen3-1.7B Wordle GRPO Training Dashboard
📈
Live training metrics for Qwen3-1.7B GRPO run on Wordle
models 13
mrinaalarora/tiny-aya-global-prudent-financial-rslora-seed0
Updated • 16
mrinaalarora/tiny-aya-global-risky-financial-rslora-seed2
Updated • 18
mrinaalarora/tiny-aya-global-risky-financial-rslora-seed1
Updated • 20
mrinaalarora/tiny-aya-global-risky-financial-rslora-seed0
Updated • 14
mrinaalarora/mrinaal-124m-instruct-v3-mathmix-smoltalk-150k
0.1B • Updated
mrinaalarora/mrinaal-124m-base-v3-mathmix
0.1B • Updated
mrinaalarora/mrinaal-124m-instruct-smoltalk-50k
0.1B • Updated
mrinaalarora/mrinaal-124m-base-v2
0.1B • Updated
mrinaalarora/mrinaal-124m-base
Updated
mrinaalarora/wordle-grpo-Qwen3-1.7B
Reinforcement Learning • 2B • Updated • 8