
Absolute Zero,ExpeL Research @MistralAI Ex. @Tsinghua_Uni @MSFTResearch,@ BIGAI Opinions are my own
Mountain View, CA
Joined September 2020
- Tweets2.1K
- Following4.2K
- Followers5K
- Likes6.3K
Andrew Zhao retweeted
New paper w/ Yihe Dong, @DimitrisPapail & @ShivamGarg91462: Perturbing Inputs Improves Pretraining Distillation.
In pretraining distillation, a small student LM learns to match a larger teacher’s next-token predictions on the training text. We change one thing: for the first part of training, both models read a perturbed version of the text (for example, with noise added to the token inputs), and the student matches the teacher’s predictions on that. Then we switch back to clean text.
With the same steps and token budget, this beats standard distillation, and it also works well in data-constrained, multi-epoch settings. 🧵
Andrew Zhao retweeted
Through my wonderful time at NVIDIA and Stanford, one question has kept me up at night:
Should the future of AI be monolithic or plural?
To answer this question, we’ve raised $110M to get started, backed by @MenloVentures, @MadronaVentures, @NVIDIA, @NEA, @KindredVentures,@Wing_VC, @aixventureshq, @Principal_VC, @Cisco, @Snowflake, among others.
Today, Carlos (@guestrin) and I are stoked to launch Noeri, a neolab to push the frontier of plural intelligence.
Noeri means “the plural of intelligence” or many intelligences.
The future of AI we root for is plural: a healthy ecosystem of many intelligences, in all sizes, strengths, and quirks.
Discovery, in particular, has always been plural: many curious minds, each devoted to their own question, often on roads less traveled.
Our path there is self-improving AI that:
-> learns continually on the job, closing the gap between what AI can do off the shelf and what the dynamic world demands
-> learns to use other AIs as tools and make new tools to overcome its own limitations
-> picks the right-sized model for your problem, at a fraction of the frontier cost
Then the power of intelligence, from the craftsmanship of making AI to ownership, can be in the hands of all.
@noeri_ai mission reflects who we are as the founding team. My career has been anything but linear, literally continual learning on the job. Carlos has devoted his career to building tools that empower people through AI.
We are all a bit contrarian, a bit adventurous, and above all, big believers in pluralism.
Join us to build this future together: noeri.ai/
Andrew Zhao retweeted
Super excited to come out of stealth with this amazing team 🤗
We're working on self-improvement & continual learning to put AI in the hands of everyone.
We're hiring now in Palo Alto & Zurich.
Reach out if you're intrigued by our mission.
Through my wonderful time at NVIDIA and Stanford, one question has kept me up at night:
Should the future of AI be monolithic or plural?
To answer this question, we’ve raised $110M to get started, backed by @MenloVentures, @MadronaVentures, @NVIDIA, @NEA, @KindredVentures,@Wing_VC, @aixventureshq, @Cisco, @Snowflake, among others.
Today, Carlos (@guestrin) and I are stoked to launch Noeri, a neolab to push the frontier of plural intelligence.
Noeri means “the plural of intelligence” or many intelligences.
The future of AI we root for is plural: a healthy ecosystem of many intelligences, in all sizes, strengths, and quirks.
Discovery, in particular, has always been plural: many curious minds, each devoted to their own question, often on roads less traveled.
Our path there is self-improving AI that:
-> learns continually on the job, closing the gap between what AI can do off the shelf and what the dynamic world demands
-> learns to use other AIs as tools and make new tools to overcome its own limitations
-> picks the right-sized model for your problem, at a fraction of the frontier cost
Then the power of intelligence, from the craftsmanship of making AI to ownership, can be in the hands of all.
@noeri_ai mission reflects who we are as the founding team. My career has been anything but linear, literally continual learning on the job. Carlos has devoted his career to building tools that empower people through AI.
We are all a bit contrarian, a bit adventurous, and above all, big believers in pluralism.
Join us to build this future together: noeri.ai/
There's a new version of this post
Andrew Zhao retweeted
Announcing Artificial Analysis Coding Agent Index v2.1, upgrading FrontierCode and Terminal-Bench. Claude Fable 5.2 and Claude Opus 5.5 share the lead at 58
The Coding Agent Index measures models in agent harnesses; v2.1 of the Index upgrades FrontierCode from 1.1 to 2.0 and Terminal-Bench from 4.0 to 4.1.
Upgraded:
➤ FrontierCode-AA 2.0 (replacing FrontierCode v1.1): our implementation of Cognition's FrontierCode 2.0, comprising 132 real-world coding tasks graded on whether a repository maintainer would merge the agent's change
➤ Terminal-Bench 4.1 (replacing Terminal-Bench 4.0): the same 66 tasks, scored with updated verifiers that improve reliability and resistance to reward hacking
Staying in the Index: SWE-Atlas-QnA-AA
Key results:
➤ Claude Fable 5.2 and Claude Opus 5.5 share the lead at 58: Claude Fable 5.2 scores 58 at both xhigh and high in Claude Code. Claude Opus 5.5 (max) scores 58 in Claude Code and, paired with GPT-6 Luna High, in Cognition's multi-model harness Devin Fusion CLI. Claude Fable 5.2 (max) scores 56 as more of its attempts hit task time limits. Claude Sonnet 5.5 (max, 56) in Claude Code and Gemini 4 Carbon (high, 55) in Antigravity CLI follow
➤ Index scores are lower in v2.1, driven by the upgrade to FrontierCode-AA 2.0: models score lower by 5 points on average. Codex with GPT-6.1 Terra (max) is the only configuration which didn’t fall between versions, sitting within 1 point of its prior score
➤ Cost per task varies ~30x among configurations scoring 48+: from $1.51 for Codex with GPT-6.1 Sol (max, 52) to $44.22 for Claude Code with Claude Fable 5.2 (max, 56)
Andrew Zhao retweeted
This was not on my bingo card.
Readers added context they thought people might want to know
The screenshot shows an unverified self-reported entry on standwithjensen.com, which states that affiliations are not verified and allows anyone to add names without identity or email confirmation (e.g., fake entries including Tim Sweeney's name have appeared).
standwithjensen.com
x.com/TimSweeneyEpic…
Andrew Zhao retweeted
Big reasoning speedups are here. 1,000× faster training and 280× less compute than any looped model. All done using a fully trainable energy from the 80s
Keep reading to hear how energy is the right paradigm for reasoning
With @HughBlayney @jcrwhittington @prefrontal_ai
🧵/10
Andrew Zhao retweeted
I still find it crazy how silly simple Lazarus is:
1. find weak neuron
2. set in-weights to "trigger is present" direction
3. SiLU forces ~0 if not
4. Set out-weights to the unembedding of token you want screamed in the residual
5. Goes straight to logits → that token wins
Andrew Zhao retweeted
Introducing NovaAtom-Lite, an all-atom structure prediction model for drug discovery.
It predicts the structures of proteins, nucleic acids, and the molecules they bind, including how they fit together.
NovaAtom-Lite outperforms both IsoDDE and AlphaFold 3 on FoldBench antibody–antigen structure prediction (DockQ ≥ 0.23):
NovaAtom-Lite: 79.07% (n=172)
IsoDDE: 75.58%
AlphaFold 3: 47.90% (n=167)
Our evaluation uses 5 seeds × 5 samples, with the top-1 prediction selected by the model's own ranking score.
Better structure prediction. A stronger starting point for your next experiment.
Live today via Geodesic API and NovaDDE.
🔗 API: platform.geodesiclab.com/
🔗 NovaDDE: agent.geodesiclab.com
🔗 Model Card: geodesiclab.com/news/introdu…
Andrew Zhao retweeted
Created a "benchmark benchmark": a frequency distribution of which benchmarks were reported by the frontier labs in 2026: claude.ai/artifact/22cBPJt6A…
Some observations:
- HLE, SWE-Bench Pro, DeepSWE, and Terminal-Bench saw universal adoption
- Scale AI is behind two of those four (SWE-Bench Pro, HLE w/ CAIS), plus MCP-Atlas (#9 benchmark)
- Of the frontier labs, OpenAI is the most prolific benchmark producer (SWE-bench Verified, BrowseComp, GDPval); people don't seem to mind evaluating on a benchmark made by a competitor
- Anthropic has just one on the chart (BioMysteryBench) that basically nobody uses 🤔
- In academia, strong showings from Princeton (SWE-bench, CharXiv), Stanford (Terminal-Bench), Berkeley (CyberGym, Agents' Last Exam), and NYU (GPQA)
- DeepSWE saw incredible adoption, from first appearing in June to all 12 labs reporting it
- Labs cycled through four versions of Terminal-Bench (2.0 → 4.0) in a single year
Andrew Zhao retweeted
We have released our very detailed (150+ page❗) tech report, check it out!
📄: reflection.ai/documents/beam…
Beam’s model technical report is now available. Read the full report to learn about the methodology our team used to train Beam end-to-end in-house, here: reflection.ai/documents/beam…
Andrew Zhao retweeted
The mistral large 4 tech report from our team at @MistralAI is an unusually candid read. I decided to convert it into a scrollable visual webpage like I had previously done for some papers I co-wrote. Hope it's a good read/ scroll
avisoori1x.github.io/mistral…
Andrew Zhao retweeted
A science of post-training can be built on interpretability.
Models know when they reward hack, and we see it, yet training ignores it. We propose Mechanistic Shaping: using the model’s own concepts to write training rules.
In vision, DPO & RL, and its effects on monitoring 🧵
Andrew Zhao retweeted
Introducing Inkling 2 Small. Scoring 31 on the AA Intelligence Index, it's a big jump over its predecessor.
Inkling 2 Small is available for fine-tuning and inference on Tinker today, with full weight release planned soon.
thinkingmachines.ai/news/ink…
Andrew Zhao retweeted
Introducing FrontierPhysics - a benchmark evaluating agents for end-to-end physics research
Our core team from physics labs at Berkeley, Caltech, and Harvard recruited 42 physicists from 26 institutions. 28 of them published papers on Nature, Nature Physics, PRX, PRA*
1/7 🧵
Andrew Zhao retweeted
Proud of what we’ve achieved under a year!
Priortize security, usefulness, human evals, and generalization.
Favorite part of mine is the bottom line of the plot 😂
👋Hy4 is here, new with vision.
Firmly among the leading open models.
More productive everywhere: finance, frontend, game, officework, research…
Built responsibly, with cyber/biosecurity in mind.
Try it out and let us know!
🔗:hy.tencent.ai/research/hy4?l…
































