arXiv:2606.18699v2 Announce Type: replace-cross Abstract: Large language models (LLMs) have shown impressive capabilities across diverse tasks, yet their…
Category: AI
A Circuit, Not The Circuit: Non-Unique Causal Localisation of the Mamba-2 State Sink
arXiv:2606.00930v2 Announce Type: replace-cross Abstract: Mechanistic interpretability routinely reads a probe and labels its top-activating units as the…
SaliMory: Orchestrating Cognitive Memory for Conversational Agents
arXiv:2606.04120v2 Announce Type: replace-cross Abstract: Conversational agents that serve as lifelong companions must maintain persistent memory across…
Google’s Gemini 3.5 Transcribe turns speech to text in 85 languages while auto-correcting your verbal stumbles
Google’s new Gemini 3.5 Transcribe recognizes over 85 languages, strips filler words, and corrects slips of the tongue in real time. It hits a 4.0 percent…
Tournament-GRPO: Group-Wise Tournament Rewards for Reinforcement Learning in Open-Ended Long-Form Generation
arXiv:2605.26958v2 Announce Type: replace-cross Abstract: Reinforcement learning in open-ended long-form generation is challenging because reliable…
Expanding OpenAI’s presence in Brazil
OpenAI is expanding its presence in Brazil, deepening engagement with developers, businesses, and communities to support AI adoption across the country.
When Can One Neuron Fix Repetition Loops in LLMs?
arXiv:2606.13705v2 Announce Type: replace-cross Abstract: The Gemma 4 instruction-tuned models share a reproducible failure: on long factual enumeration…
Anthropic locks in 45-billion-dollar compute deal with Nscale ahead of IPO
Anthropic has struck a deal worth around 45 billion dollars with British cloud startup Nscale, Bloomberg reports. The article Anthropic locks in…
Skill-Conditioned Gated Self-Distillation for LLM Reasoning
arXiv:2605.28791v2 Announce Type: replace-cross Abstract: On-policy self-distillation (SD) improves LLM reasoning by using teacher-side privileged…
ForceFlow: Learning to Feel and Act via Contact-Driven Flow Matching
arXiv:2605.11048v2 Announce Type: replace-cross Abstract: Existing imitation learning methods enable robots to interact autonomously with the physical…