PoliticsBench: Benchmarking Political Values in Large Language Models with Multi-Turn Roleplay · HackerPapers discussions

PoliticsBench: Benchmarking Political Values in Large Language Models with Multi-Turn Roleplay — mentioned by dijit

Thu, 09 Jul 2026 06:56:26 GMT

These articles are what makes me say that, https://arxiv.org/pdf/2603.23841 https://www.washingtonpost.com/technology/interactive/2026/0... (X post about it: https://x.com/genejchan/status/2070099701547090340) https://www.promptfoo.dev/blog/grok-4-political-bias/

Paper and discussion history. A mention is not an endorsement.

PoliticsBench: Benchmarking Political Values in Large Language Models with Multi-Turn Roleplay — mentioned by FunHearing3443

Wed, 08 Jul 2026 22:36:58 GMT

Studies on political bias in models consistently show that LLMs lean politically left. The only outlier is grok which leans right but by a smaller factor, according to this study for instance: https://arxiv.org/abs/2603.23841 Edit: adding some other studies that are easily retrievable with a quick search for those unsatisfied with the first one - https://arxiv.org/abs/2606.12922 https://arxiv.org/abs/2412.16746 Claiming reality has a left wing bias is certainly an opinion you're welcome to have to explain this, but the reality of the bias in models is well evidenced. It seems that practically…

Paper and discussion history. A mention is not an endorsement.

PoliticsBench: Benchmarking Political Values in Large Language Models with Multi-Turn Roleplay — mentioned by WatchDog

Wed, 01 Apr 2026 23:59:36 GMT

Overall it's worse than the other frontier models, but it's decent for queries about breaking news, due to being trained on twitter data. It's also better for queries about controversial topics, and topics that the other labs have deemed to be "unsafe". Politically, it differs quite a bit from other models.[0] It's right leaning, although it's closer neutral than other models, defining what neutral is a challenge though. [0]: https://arxiv.org/abs/2603.23841

Paper and discussion history. A mention is not an endorsement.