[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

![omarsar0 Avatar](https://lunarcrush.com/gi/w:24/cr:twitter::3448284313.png) elvis [@omarsar0](/creator/twitter/omarsar0) on x 254.7K followers
Created: 2025-07-14 15:17:07 UTC

Overview

Investigates the surprising fragility of LLM-based reward models used in Reinforcement Learning with Verifiable Rewards (RLVR).

The authors find that inserting superficial, semantically empty tokens, like “Thought process:”, “Solution”, or even just a colon “:”, can consistently trick models into giving false positive rewards, regardless of the actual correctness of the response.

![](https://pbs.twimg.com/media/Gv09dR7aIAACHp_.jpg)

XXXXX engagements

![Engagements Line Chart](https://lunarcrush.com/gi/w:600/p:tweet::1944778190695940448/c:line.svg)

**Related Topics**
[elvis](/topic/elvis)

[Post Link](https://x.com/omarsar0/status/1944778190695940448)