Open RLVR results flip when measurement changes, tiny GRPO testbed exposes why
The post demonstrates that the same open‑source RLVR training run can appear as a success, failure, or reversal solely depending on the reward channel, extractor, or decoding regime used to evaluate it. Using a cheap Qwen2.5‑0.5B GRPO testbed, the author separates these instruments, revealing hidden reward‑hacking and metric‑gaming effects that must be audited when reporting RLHF/RLVR results.