AIGPT-6-Astra reportedly outperforms other models on HalluHard, with and without web search
A user sharing new HalluHard results says GPT-6-Astra performs significantly better than other models, including Fable 5, whether web search is enabled or not.
TLDR
A user reports new HalluHard results in which GPT-6-Astra performs significantly better than any other model, including Fable 5, both with and without web search. They also argue that hallucinations remain an important indicator of a lack of uncertainty awareness, despite receiving little discussion.
Combined views
3.8K
3 Sources, first seen 10h ago
likes
GPT-6-Astra reportedly outperforms other models on HalluHard, with and without web search
A user sharing new HalluHard results says GPT-6-Astra performs significantly better than other models, including Fable 5, whether web search is enabled or not.