But are there currently any good local models that can provide good conversation/support of a recent paper? I feel the level of hallucinations will be huge. Eve

But are there currently any good local models that can provide good conversation/support of a recent paper? I feel the level of hallucinations will be huge. Even with gpt4 with pdf2text, it tends to hallucinate a lot. And Gpt4 is the best LLM model so far.