Downloads · 30 days
0
Shaer-AI-2/Shaer-adapters-grpo-short1k-no-trio-v2
Shaer-adapters-grpo-short1k-no-trio-v2 is a machine learning model from Shaer-AI-2. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for transformers. The card lists the license as other.
This repo is the current best completed GRPO stage in the Shaer project.
Downloads · 30 days
0
Access
Public
Updated Apr 13, 2026
Repo size
188 GB
Likes
0
Public
Click a slice to open those files.
.safetensors2.6 GB · 43%
From the Hugging Face model README
This repo is the current best completed GRPO stage in the Shaer project.
This is still the best completed GRPO run in the project.
Later runs v3 and v4 were useful diagnostically, but neither surpassed v2 as a finished model. The current unlaunched next-step candidate is the dual-judge v5 setup prepared locally in /root/workspace/Shaer/grpo.
Project sequence:
Shaer-AI/Shaer-adapters
trusted clean SFT baselineShaer-AI/Shaer-adapters-grpo
historically important but reward-hacked GRPO stageShaer-AI/Shaer-adapters-grpo-vnext
first strict anti-hack rerunShaer-AI/Shaer-adapters-grpo-friend-v1
first judge-centered GRPO runShaer-AI/Shaer-adapters-grpo-friend-v1-easyfirst
healthier multiplicative judge-centered rerunShaer-AI/Shaer-adapters-grpo-short1k-no-trio-v2
smoother weighted-reward rerun on a short easy-first subsetShaer-AI/Shaer-adapters-grpo-short1k-no-trio-v3
meter-times-judge-core rerun on the same short subsetThis repo is the first stage where the project adopted both:
Shaer-AI/Shaer-adaptersShaer-AI/ashaar-enhanced-desc-baseform-final-sft-lte20-min500-splits-grpo-meter-count-v1Shaer-AI/ashaar-with-enhanced-descriptions-baseform-final-sft-lte20-min500-splits1-6 baytsالمديد, المنسرح, الهزج1000 rows per surviving meter995580 rows total/root/workspace/Shaer/grpo/outputs/train/shaer_grpo_20260412_221146This run introduced the weighted reward:
reward_train = hard_gate * (
0.45 * meter
+ 0.15 * count_adherence
+ 0.30 * judge_quality
+ 0.10 * repeat_soft
)
Key ideas:
Judge setup in this stage:
checkpoint-32000.78620.93950.44450.99190.90040.9875Final eval at step 3300:
0.77280.92510.42100.99750.90310.9750This is the healthiest completed GRPO regime so far in the project:
But this is still not the final flagship paper model:
Manual inspection of this run made the next issue unusually clear:
That led directly to the next stage:
Shaer-AI/Shaer-adapters-grpo-short1k-no-trio-v3whose main change is:
reward_train = hard_gate * (
0.65 * (meter * judge_quality)
+ 0.20 * count_adherence
+ 0.15 * repeat_soft
)
The whole point of that next run is to make semantic weakness unable to hide behind great meter.
/root/workspace/Shaer/grpo/outputs/train/shaer_grpo_20260412_221146eval_manual_inspection.mdrun_summary.jsonUse this repo as the current best completed GRPO stage in the Shaer project.
For paper writing, present it as: