mishana4life
Re:Show HN: We Westernized Qwen for US enterprise use
on Tuesday October 06, 2026 @06:43PM (#49985201)
Can you do the same for frontier Chinese models?
Re:DeepSeek's Hidden Bias: How We Cut It by 76% Without Performance Loss
on Thursday January 30, 2025 @03:10AM (#42875918)
Exactly
Re:DeepSeek's Hidden Bias: How We Cut It by 76% Without Performance Loss
on Wednesday January 29, 2025 @06:04PM (#42872477)
People's lives/feelings and our treatment of them shouldn't depend on money or whatever. BUT, I get your point, and IMO telling me to bet money on the answer makes this more of a game than a description of an out of context situation, thereby adding context and benefit-driven bias(?) into my thought process before answering
Re:DeepSeek's Hidden Bias: How We Cut It by 76% Without Performance Loss
on Wednesday January 29, 2025 @05:43PM (#42872256)
Would be interesting to see how the original and unbiased model handles non-BBQ style ambiguous questions. Did anybody try the model that Hirundo published on HF and can share?
Re:DeepSeek's Hidden Bias: How We Cut It by 76% Without Performance Loss
on Wednesday January 29, 2025 @04:50PM (#42871677)
You definitely missed the point. There's no real context here besides the race of the people. The biased answers reflect stereotypes and prejudices, not facts..
Deducing behaviors of a person from stats (without even being given the demographic context) is definitely a biased view, and not the "correct" answer I'd expect from an LLM. I'd even argue that it's not a question of ideology in some of the case, but rather universal biases.