Basis

Basis

[Basis @BasisOrg]

We're at ICML this week presenting work on world models, dynamical systems modeling, and more. If you're here and would like to discuss over a drink, come to our social on Thursday evening! Link below.

[Basis @BasisOrg]

Basis researcher @ralph_ep will be in Texas this week to give a talk at PestWorld 2026! 🐀 Catch his talk, "Eavesdropping on Urban Rats: Bringing AI and Field Neuroscience to Rodent Monitoring and Mitigation," on Thursday 10/22 from 1:30-2:45 PM in Fort Worth 6-7.

[Basis @BasisOrg]

Most programming languages lack the precise rules software verification requires—a growing bottleneck as AI gets better at writing proofs. @kirancodes used LLMs to write those rules for a small language and proved in Lean that they match how the code runs. https://t.co/GTNfsF0Xux

[Zenna Tavares @ZennaTavares]

For a while now, @kirancodes has been excitedly telling everyone at @BasisOrg who would listen about a crazy idea she's been pursuing. Not everyone thought it was possible, but it's starting to work. See https://t.co/A2FyTJzSsA

TL;DR: The bottleneck in verifying the world's software is shifting from theorem proving to specifications and semantics. We now have a way to verify semantics that can scale across all programming languages.

The basic idea:

To verify that a program is correct, you need the program, a specification you want to prove, and a semantics that defines what programs and specs mean.

Most programming languages (Python, Ruby, Lua…) have no formal semantics, because semantics are painful to write for complex languages.

You can ask an LLM to write a semantics for you, but it'll probably have mistakes.

But (i) programming languages do have interpreters/compilers, which ultimately are compiled down to machine code, and (ii) some instruction sets do have formal semantics. RISC-V has an official formal model.

So, crazy idea:

Step 1: Compile your interpreter (e.g. Python's) down to RISC-V. Now its behaviour is formally defined. It's exact, but unreadable: thousands of machine instructions.

Step 2: Ask your LLM to write a semantics for your language.

Step 3: Ask the LLM to prove, in @leanprover, that its semantics agrees with the compiled interpreter on every possible program. If the semantics is wrong, the proof gets stuck, and the LLM goes back to fix it.

Step 4: If the proof goes through, you have a semantics you don't need to trust, and you can verify programs in a language that never had one, with guarantees that reach the binary that actually runs.

Kiran has now done this end to end for a simple WHILE language, and we're scaling it up to Lua and OCaml, with Python as the north star.

[Basis @BasisOrg]

@wbz featured Basis researcher Nick Jourjine yesterday discussing our rat research in Boston! We’re using thermal video, acoustic monitoring, and AI to study urban rats’ behavior and help cities manage rat populations humanely.

Watch: https://t.co/SGMCz9Yb7g

[Basis @BasisOrg]

How do urban rats and other animals move around and socialize in the wild? 🐀

At Basis, we’re developing tools to study these questions using thermal video, 3D reconstructions, and synthetic data.

Learn more here: https://t.co/7xxL6Nkyja

[Basis @BasisOrg]

Last week Basis researchers attended the National Urban Rat Summit, where @Ralph_Ep presented our work using acoustic monitoring to study rat behavior, while Nick Jourjine showcased monitoring devices on a nighttime “rat walk” 🐀.
Ralph and @e_Mackevicius also met @MayorWu. Ralph spoke with WBUR about the research: https://t.co/wLq4IIUen0

[Zenna Tavares @ZennaTavares]

I'm at #ICML2026 through Sunday with a small cohort from @BasisOrg . Come find us to discuss our projects (world model learning, robot design, formal verification, urban ecology), and our general reasoning systema beneath them

And: event on Thursday evening at 7. Link below

[Basis @BasisOrg]

See event details and RSVP: https://t.co/UWrokm4ixP

[Basis @BasisOrg]

We're at ICML this week presenting work on world models, dynamical systems modeling, and more. If you're here and would like to discuss over a drink, come to our social on Thursday evening! Link below.

[Basis @BasisOrg]

We're hiring research scientists in PL and other areas. Join us! https://t.co/ahyAqxkmQY

[Basis @BasisOrg]

We're attending and sponsoring #POPL2026 in Rennes, France 🇫🇷 -- if you're around, stop by our sponsor booth to chat about research and open opportunities at Basis. We'll be there Wednesday 10:00-19:30 and Friday 10:00-18:00.

[Yichao Liang @yichao_liang]

New preprint on learning abstract world models for robotics planning. Paper + code below. 🤖🌐

Must an agent plan by simulating pixels frame by frame, or can it think in abstractions? Consider planning an international flight: we can reason about buying tickets, changing airplanes, and crossing borders without committing to the color of the airplane or the milliseconds before takeoff. Absent abstraction, planning over long time horizons would be intractable, because every minute detail of the world would need to be simulated.
[1/7]

[Basis @BasisOrg]

@alex_prompter To everyone reading: Basis is hiring! Join us! https://t.co/N75gfscXt5

[Basis @BasisOrg]

@alex_prompter Thank you for the thread! Come discuss at our NeurIPS social if you'll be around https://t.co/8HzHjTUbKo

[Alex Veremeyenko @alex_verem]

🚨 MIT and Basis Research just dropped a new way to measure if AI actually understands the world and the results are brutal.

It’s called "WorldTest", and it doesn’t just check how well an AI predicts the next frame or maximizes reward.

It checks whether the model can build an internal model of reality and use it to handle new situations.

They built 'AutumnBench', a suite of 43 interactive worlds and 129 tasks where AIs must:

• Predict hidden parts of the world (masked-frame prediction)
• Plan sequences of actions to reach a goal
• Detect when the environment’s rules suddenly change

Then they tested 517 humans vs. top AI models Claude, Gemini 2.5 Pro, and o3.

Humans crushed every model. Even massive compute scaling barely helped.

The takeaway is wild... current AIs don’t understand environments; they pattern-match inside them.

They don’t explore strategically, revise beliefs, or run experiments like humans do.

WorldTest might be the first benchmark that actually measures understanding, not memorization.

The gap it reveals isn’t small it’s the next grand challenge in AI cognition.

Paper: Benchmarking World-Model Learning (arxiv. org/abs/2510.19788)

[Gary Marcus @GaryMarcus]

like i have been saying since 2019, world models are the next key step.

[Eric Bourdages @EZE3D]

"Today’s AIs don’t understand environments; they just pattern-match inside them."

Literally what critics have been saying for years now.

[Basis @BasisOrg]

We'll also be at NeurIPS; come talk to us! Visit our booth or register for our social: https://t.co/8HzHjTUJzW

[Basis @BasisOrg]

New paper from Basis' Project MARA team and collabs. The ability to learn and use world models is a key aspect of human intelligence, but evaluating this ability remains elusive. In this work we propose WorldTest, a representation-agnostic, behavior-based agent eval framework.

[Basis @BasisOrg]

All open roles: https://t.co/KLWI31J4Zq