Search papers, labs, and topics across Lattice.
4
3
6
The DeepSeek-V4.1-Flash model, a multimodal Mixture-of-Experts model with 552B backbone parameters and support for contexts of up to one million tokens, is introduced, substantially improving cost efficiency for agentic workloads and pushing the limits of KV cache compression.
RING achieves superior knowledge integration without the latency of external retrieval, redefining efficiency in large-scale language models.
Overthinking in RL models can amplify during training, but Dynamic Rollout Editing effectively curtails this behavior, enhancing reasoning efficiency.
Latent reasoning can now outperform explicit reasoning in complex tasks, thanks to a new RL method that stabilizes training by explicitly handling issues like invalid latent states and misaligned token-level updates.
