AI Model2026
A sparsely activated Mixture-of-Experts (MoE) causal language model with 16B total parameters and 2.8B active parameters per token, released with end-to-end checkpoints and training recipe; trained on AMD Instinct GPUs and licensed for research use.
AI Agent Papers2026
Alternates targeted research and constraint-wise audits to recursively improve long-horizon answers: an inner loop gathers evidence and drafts solutions, an outer loop audits unresolved claims and launches focused follow-ups. Trains 4B dense and 122B-A10B MoE agents with long-horizon RL and agentic mid-training, outperforming comparable-scale baselines on multi-step research benchmarks.
AI Model2026
A text-only open-weight MOE code model (35B total, 3B active) fine-tuned with SFT+RL for agentic coding; achieves strong agentic-code benchmarks, supports 262k context and deployment via Transformers/vLLM; vision weights are not included.
Presents Skill Self-Play (Skill-SP), a co-evolutionary training loop where a proposer, solver, and dynamic skill controller generate, solve, and verify tasks conditioned on reusable skills — balancing verifiable execution with open-ended task diversity to boost LLM tool-use and reasoning.
Enables tactile-aware robot manipulation by pretraining a vision–tactile–language–action foundation model and improving offline policies with ALTER. Combines large-scale NeoData visuo-tactile pretraining, a latent tactile pathway for predictive touch signals, and advantage‑conditioned offline RL for contact-rich tasks.
Presents a 2.8T-parameter Mixture-of-Experts multimodal model with a 1-million-token context window and 104 billion activated parameters, targeting long-horizon agentic RL, coding, reasoning, and vision. Key innovations include Kimi Delta Attention, Attention Residuals, Stable LatentMoE (16 of 896 experts active per token), ~2.5× scaling efficiency over Kimi K2, and a public weight release.
AI Agent Papers2026
Bridges the proprietary-to-open-source gap in agentic search by converting multi-step retrieval and reasoning traces into a structured, style-normalized JSON protocol and using it for joint distillation + RL. Produces denser supervision that improves student success rates while reducing style drift.
Allocates token-level credit in rubric-conditioned GRPO by counterfactually replaying the same response under rubric and criteria-free prompts, using tokenwise log-likelihood contrasts to compute bounded, response-normalized weights that redistribute GRPO advantages without training an auxiliary scorer.
Autonomously proposes, modifies, executes, and evaluates ML experiments to study recursive self-improvement in machine learning engineering. Implements an open stack (OpenMLE-Gym, -RL, -Evo) and post-trains Frontis-MA1 (35B) around four evolution operators (Draft, Improve, Debug, Crossover); releases model weights and the full codebase.
AI Agent Papers2026
Designs and evaluates a foundation GUI agent that performs cross-platform GUI and CLI actions on real devices to complete long-horizon workflows. Emphasizes a unified action space, a large-scale real-device mobile runtime, an AutoResearch-style data flywheel, and online RL training across 10,000+ concurrent environments.
Regularizes latent world models by replacing the Epps–Pulley Gaussianization objective with a quantile–quantile matching loss that aligns projected latent samples to rank-matched Gaussian quantiles, improving tail correction and planning success via cross-batch ranking.
AI Agent Papers2026
Evaluates vision-language model judges on computer-using agent (CUA) trajectories to measure verifier reliability. Provides OSReward-Hard and OSReward-Multi challenge sets, the OS-Shepherd-100K reasoning-annotated corpus, and trained OS-Shepherd reward models that match commercial judges at ~30–60× lower cost.