tinker-rl-bench-scale_gsm8k_llama-8b-inst – AI Model by arvindcr4 | AlphaNeural AI

You can deploy this model and start earning money today!

arvindcr4/
tinker-rl-bench-scale_gsm8k_llama-8b-inst

like

0

reinforcement-learning

tinker-rl-bench

meta-llama/Llama-3.1-8B-Instruct

Views

No views yet

scale_gsm8k_llama-8b-inst

GRPO experiment from TinkerRL-Bench world-class experiment suite.

Training Details

Base model: meta-llama/Llama-3.1-8B-Instruct
Method: GRPO (Group Relative Policy Optimization)
Platform: Tinker API v0.18.1
Task: gsm8k
Seed: 42
LoRA rank: 32
Learning rate: 3e-05
Group size: 8
Steps: 30

Results

First-5 avg reward: 97.5%
Last-10 avg reward: 84.4%
Peak reward: 100.0%
Zero-loss steps: 43%
Tinker Run ID: 488bbb2e-aa35-5431-b4a5-793f68b6bc49:train:0

Reward Trace

[
  1.0,
  0.875,
  1.0,
  1.0,
  1.0,
  0.625,
  1.0,
  1.0,
  0.875,
  0.875,
  1.0,
  1.0,
  0.9375,
  0.75,
  1.0,
  0.375,
  0.8125,
  0.875,
  0.875,
  0.75,
  0.875,
  0.6875,
  0.75,
  0.4375,
  1.0,
  1.0,
  0.8125,
  1.0,
  0.875,
  1.0
]

Citation

bibtex

@misc{tinker-rl-bench-2026,
  title={TinkerRL-Bench: A Unified Benchmark for RL Post-Training},
  author={Arvind C R and Sandhya Jeyaraj and Madhu Kumara L and Mohammad Rafi and Dhruva N Murthy and Arumugam K},
  year={2026},
  url={https://github.com/arvindcr4/tinker-rl-lab}
}