Request access to CausalOPD checkpoints
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
These checkpoints are released for non-commercial academic research only. Requests are reviewed manually. Only applicants using an institutional academic e-mail address (e.g. *.edu, *.edu.cn, *.ac.uk) will be approved; requests from personal e-mail providers will be declined.
Log in or Sign Up to review the conditions and access this model content.
CausalOPD checkpoints
Training outputs (model weights, configs, trainer states and logs) from the CausalOPD experiments (Qwen3-8B / Qwen2.5-7B / OLMo-3-7B students; science and tool-use settings).
Optimizer / scheduler / RNG states are not included, so runs cannot be resumed bit-exactly, but every saved checkpoint's weights are provided.
Layout mirrors the original outputs/ tree, e.g.
exp_science_native_qwen3_8b/run/native_qwen3_8b_our_method_direct_coef8_seed42/checkpoint-668/.
Access is gated with manual approval; see the request form above.