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.
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.
AI Agent Papers2026
A 365-day, order-level simulation benchmark for evaluating long-term coherence of LLM agents in seller-side e-commerce. Grounded in 98,843 real product records and 26 interactive tools, it pairs prompt upstream supplier signals with delayed downstream order outcomes to stress planning, memory, and tool use over long horizons.
AI Model2026
An OpenAI-compatible LLM checkpoint optimized for agentic and long-context scenarios, shipping DSpark speculative decoding and vLLM/SGLang deployment recipes; tailored for code-agent and multi-step reasoning workloads and released under MIT.
AI Agent Papers2026
Treats agent self-improvement as natural selection over a population of harnesses (prompts, tools, skills, control flow), evolving a frozen-model agent by selecting harness edits that extend capability without regressing others. Uses a preserve-and-extend contract, lineage archive, and verifier-driven fitness (no gold solutions) to recombine complementary edits and transfer across benchmarks.
AI Model2026
Enables local use of a GGUF-quantized DeepSeek-V4-Flash-0731 via Unsloth Dynamic quantizations; provides a Q8 (162GB) lossless option and smaller Q4 variants for lower-memory inference and agentic scenarios using Unsloth tooling.
AI Model2026
Finetunes Qwen3.6‑35B using an adversarial generator–critic loop that synthesizes ~10,000 verifiable “frontier” tasks to boost scientific research, long‑horizon reasoning, coding and tool use; supports an extended 262,144‑token context and common serving stacks.
AI Model2026
A 35B additive ternary-quantized variant of Mach-1 that aims to preserve most capabilities of its BF16 teacher while reducing compute and memory; shows ~95% mean retention across 12 benchmarks and competitive per-task parity on several evaluations.
A pretrain-then-transfer method for streaming recommendation that decouples refreshable behavioral knowledge from task-specific geometry to enable continual model refresh without downstream interference; introduces Behavioral Multi-Token Prediction and Anchored Calibration Residual and shows 4–12% offline gains plus live Shopee A/B lifts.
Frames skill generation as a sequential editing task and introduces a novel rollback reward to train an RL generator (Skill-α) that evaluates each edit by its downstream execution impact, producing skills that improve agent success rates across document-to-skill and experience-to-skill settings.
AI Video Papers2026
Introduces WorldExam, a diagnostic benchmark that evaluates controllable video world models across four levels from visual quality to inherent world reactivity. Covers 1,474 cases across eight tasks and supports camera-, action-, and language-driven paradigms, measuring scene-conditioned reactions beyond explicit instructions.