Shi

Webpage: praxis-research.org/grafting | Paper link: https://arxiv.org/abs/2610.00767 Work done during Peter and Dani's MATS Fellowship (10.0) under the mentorship of Shi, with Jinghua Ou as Research Manager. TLDR; We share a simple modification to synthetic document fine-tuning (SDF) that works surprisingly well: do fine-tuning on a pre-training checkpoint and add the...

Code: https://github.com/atagade/sgtr-em Model checkpoints: https://github.com/atagade/sgtr-em/blob/main/MODEL_README.md Full paper (v2 coming soon): https://arxiv.org/abs/2606.23700 TL;DR * A model’s self-model is its representation of itself. We operationalize it two ways: self-recognition (distinguishing its own outputs from others') and self-report (what it says about itself when directly asked). Interventions on either modulate Emergent Misalignment (EM)....

We pose the following research question: how can we measure the "research taste" of language models in experiment planning? What parts of planning taste remain intrinsic to humans? TL;DR. A future where “tasteless autoresearch” improves capabilities but not safety is plausible and dangerous. We need rough tests of tasteful planning...

The first three sections are written for a general TAIS reader who wants to understand what the state of Debate research is and some high-level takeaways of our work. A reader familiar with Debate may like to skip the setup and start with our presentation of An illustrative training run....

TL;DR: * Safety assurance (safety cases, heuristic arguments) involves many interpretive questions about the model’s mechanisms (e.g., motivation). These are fundamental ML problems with a new layer of difficulty due to conceptual ambiguity. * Adversarial robustness and non-human-likeness of models create epistemic challenges in our empirical investigation of interpretive questions:...

Note: this post is an explainer for our recent paper, as well as an update our previous blog post. The work was done by David, Rustem and Taywon under the mentorship of Shi Feng during MATS 9.1, with research management by Jinghua Ou. Scheming or performative scheming? (This section introduces...

This work was done by Rustem Turtayev, David Vella Zarb, and Taywon Min during MATS 9.0, mentored by Shi Feng, based on prior work by David Baek. We are grateful to our research manager Jinghua Ou for helpful suggestions on this blog post. TL;DR: in this construct validity exercise, we...