2/ We developed an algorithm called Mulberry, that aims to cover tasks in a domain while also increasing their difficulty - named after temporary portable harbours from WWII @harborframework
3/ Read more in our blog post where we explain how tasks were generated and how they provide a learnable signal for post-training: vmax.ai/team/modal-coder-sel….
Vmax retweeted
The first AI persona you’ll actually want to meet face-to-face.
Most AI personas stare blankly while you talk. Emma reacts while you’re still speaking.
After 18 months of research at @nuance_ai, this is Emma, built on Nuplex, our full-duplex audiovisual foundation model.
Vmax retweeted
while working on agentick (to be presented at NeurIPS'26), @creus_roger found that ASCII inputs seemed to really help LLMs
during his internship at @VmaxAI he pushed on this idea and came up with GlyphBench, which serves as a benchmark, a research tool, & post-training suite!
10/ 🔎 How does an agent arrive at its next move?
Our replay UI brings observations, reasoning, actions, rewards, and memory into one view.
Step through a trajectory to inspect decisions and understand where an agent succeeds or gets stuck.
ALT GlyphBench replay UI showing a Miniatari BankHeist trajectory with the glyph observation, task, reasoning, chosen action, and environment feedback.
1
10
302
11/ Led by @creus_roger, in collaboration with @Cote_Marc, @GlenBerseth and @pcastr at @Mila_Quebec & @UMontreal.
📄 Paper: arxiv.org/abs/2609.34214
💻 Code: github.com/VmaxAI/glyphbench
🌐 Blog: vmax.ai/team/glyphbench
ALT Paper title, six authors and their Mila, Université de Montréal, and Vmax affiliations, followed by the abstract.
Vmax retweeted
We built cybersecurity monitors for Kimi K3 and GLM 5.3 — more accurate, 50x faster, & 50x cheaper than an optimized LLM judge.
External red-teaming from @farairesearch found that our monitor greatly reduces universal jailbreaks. 🧵
Vmax retweeted
Come join us next week at our CoLM workshop:
Context Beyond the Window
We have some great speakers coming to speak on the opportunities, limitations and safety-considerations of current LLMs.
We will also be cohosting a social alongside Reflection!
luma.com/eocj5cm5
Vmax retweeted
Want to train your SWE agents 2× faster 🏎️💨?
We introduce KV-streams, which, rather than flushing the KV cache on agentic compaction, we preserve it and keep generating. This avoids the re-prefill cost on both the inference engine and the trainer.
Unlike other works that assume a fixed compaction strategy, we show that KV-streams can work with any compaction method in agentic tasks.
Using what we learn from experiments in text-based games. We scale KV-streams to SWE tasks, where we match the performance of re-prefill compaction and full context in half the time 🚀
arxiv.org/pdf/2609.35750
🧵
