Piaget, a language model finetuned on 15k psychological and philosophical reasoning traces.
Piaget is based on Qwen3 and was finetuned on a subset of open reasoning traces from
Dolphin R1 and
General Reasoning.
1from transformers import AutoTokenizer, AutoModelForCausalLM
2from transformers.pipelines import pipeline
3import torch
4
5repo = "gustavecortal/Piaget-4B"
6tokenizer = AutoTokenizer.from_pretrained(repo, trust_remote_code=True)
7model = AutoModelForCausalLM.from_pretrained(
8 repo, torch_dtype=torch.bfloat16, device_map="auto", trust_remote_code=True
9)
10
11pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
12
13prompt = tokenizer.apply_chat_template(
14 [
15 {
16 "role": "user",
17 "content": "Create a new psychotherapeutic technique based on cybernetic principles",
18 }
19 ],
20 tokenize=False,
21 add_generation_prompt=True,
22 enable_thinking=True,
23)
24
25print(pipe(prompt, max_new_tokens=2048, do_sample=True)[0]["generated_text"])
We performed domain filtering on
Dolphin R1 and
General Reasoning.
Prompts were embedded, clustered with k-means (k=20 000) and majority-voted for domain labels using
Qwen3-1.7B, following the
Intelligent Internet pipeline.
Clusters tagged psychology or philosophy were retained for LoRA finetuning (rank=8, alpha=16, max length=2048, epoch=1, batch size=16).
This work was performed using HPC resources (Jean Zay supercomputer) from GENCI-IDRIS (Grant 20XX-AD011014205).
Piaget aims to reason about psychological and philosophical concepts such as self-image, emotion, and existence.
