Xijia (Ray) Che

Working on AI and Security. CS Ph.D. student @UofIllinois, B.S. & M.E. @Tsinghua_Uni.

[christine @christinecorryy]

In case you need more confirmation that GPT-Astra is terrifyingly good at no-CoT computation: I ran my replication of Ryan Greenblatt’s no-CoT evals on Astra (+ Gemini 3.1 Pro, Kimi k3, Fable 5.1). It is a qualitative jump on every dataset, especially competition math and multi-hop reasoning. 🧵

[Xijia (Ray) Che @RayXijiaChe]

Excited to share that I'll be attending Context Week at Berkeley! Looking forward to connecting with others working on AI safety and security, and diving deeper into the space.

Link: https://t.co/y0pBbX0Mxu

#AISafety #ContextWeek #Berkeley

[Andrew Ng @AndrewYNg]

New: A map of the most important skills in AI Engineering. https://t.co/VVkn1Dqp1N

[Yixiong Hao @Yixiong_Hao]

Come work on replications of AI safety research with @zephaniahbroe and I at @SPARexec! @secondlookxlab is hosting 2-4 mentees over the Fall semester.

SPAR runs Sept-Dec, part time, and our application should take <1 hour. You can expect 2+ hours of mentorship from Second Look along with external mentorship from researchers at orgs like Anthropic, X AI, Oxford, Redwood Research, and Geodesic.

https://t.co/TfMjsFeBtT

[Xijia (Ray) Che @RayXijiaChe]

OpenAI now has watermarks in audio! Great step. Hoping to see watermarks in more scenarios.

[Pushmeet Kohli @pushmeet]

Great to see @OpenAI join @ElevenLabs in adopting @Google's SynthID for audio, enabling their users to have access to the same robust, imperceptible safeguards that power our own products.

For years, we at @GoogleDeepMind have been pioneering research on SynthID watermarking to address the risks of AI-generated media (particularly acute for audio and voices).

We had successfully integrated SynthID to safeguard our product surfaces - Gemini Live, Lyria, and Veo, but protecting the ecosystem requires an industry-wide effort to build foundational safety infrastructure together which is now gaining momentum!

SynthID: https://t.co/XqcVCcj9sb

OpenAI Verification Portal: https://t.co/sLgYvYIo12

ElevenLabs: https://t.co/Dbfhgq0axk

[Pushmeet Kohli @pushmeet]

Great to see @OpenAI join @ElevenLabs in adopting @Google's SynthID for audio, enabling their users to have access to the same robust, imperceptible safeguards that power our own products.

For years, we at @GoogleDeepMind have been pioneering research on SynthID watermarking to address the risks of AI-generated media (particularly acute for audio and voices).

We had successfully integrated SynthID to safeguard our product surfaces - Gemini Live, Lyria, and Veo, but protecting the ecosystem requires an industry-wide effort to build foundational safety infrastructure together which is now gaining momentum!

SynthID: https://t.co/XqcVCcj9sb

OpenAI Verification Portal: https://t.co/sLgYvYIo12

ElevenLabs: https://t.co/Dbfhgq0axk

[christine @christinecorryy]

Can frontier models solve hard problems in a single forward pass without thinking? For my first project w/ @secondlookxlab I extended Ryan Greenblatt's no-CoT evals to Fable 5, Opus 5, and GPT-5.6-Sol and saw a substantial jump on some evals!

[Xijia (Ray) Che @RayXijiaChe]

This is a work by @secondlookxlab, supervised by @zephaniahbroe. For more information, read the full post on Lesswrong: https://t.co/T46wW7pcgO

[Xijia (Ray) Che @RayXijiaChe]

Why do adversarial examples exist? Recent research suggests that adversarial attacks leverage interference between features. We replicated superposition in toy models, and found that PGD attacks align with the optimal perturbation direction predicted by latent geometry.

[Xijia (Ray) Che @RayXijiaChe]

Further, can superposition explain why adversarial examples transfer across independently-trained models? Yes, and here is the logic chain we proved: data feature correlation---&gt; interference geometry convergence ---&gt; adversarial attack transfer.

[Finn Cairns @cheeetoo_]

Can LLMs control their internal states? Anthropic found Claude can “think about” a concept on command without saying it. We replicated this in 14 open-weight models and found it to be a general property, maybe not ‘emergent’, appearing in models even as small as 270M parameters.

[Arav Dhoot @AravDhoot]

Read the work here: https://t.co/7QGq4zbWK2

Shout out to @Yixiong_Hao, @zephaniahbroe, and @hersheysbaklava for their help!

[Daniel Kokotajlo @DKokotajlo]

In AI 2027, we predicted that AI would take over the world or irreversibly concentrate power.

In AI 2040: Plan A, we've laid out our positive vision for what should happen instead.

[Arav Dhoot @AravDhoot]

In my first project for @secondlookxlab, I replicated and dug deeper into Google Deepmind's "When Chain of Thought is Necessary, Language Models Struggle to Evade Monitors".

[Xijia (Ray) Che @RayXijiaChe]

Just posted my first blog on substack!

Time Poverty Is the Greatest Poverty https://t.co/ZFoinn8xj8

[Xijia (Ray) Che @RayXijiaChe]

Just started the Second Look Fellow program at UChicago! https://t.co/Qn3pzQHen5

I will be in Chicago for the summer. Looking forward to coffee chats!

[Xijia (Ray) Che @RayXijiaChe]

Before bed, I asked Claude to queue the next 10 hours of work. It replied, “Sleep well—goodnight—I’ll keep watch.”

Moments like this hits me that the future has arrived—and that the meaning of life and one's own value need to be re-anchored in this new era.

[Xijia (Ray) Che @RayXijiaChe]

At the time of LLM, this is a talk that every security researcher should watch.

Nicholas Carlini - Black-hat LLMs | [un]prompted 2026 https://t.co/uM0zWGayxB via @YouTube