Tag
Explore by tags
An FP8-quantized, uncensored mirror of Qwen3.8-27B for image-text-to-text tasks — preserves native multimodal vision and very long context while targeting transformers/vLLM deployments; intended for offline testing and red-teaming and may bypass built-in safety filters.
Evaluates and trains multimodal agents to construct interactive 3D open worlds from user queries — provides a large benchmark of assets, seed worlds, and reverse-synthesized queries plus a sandbox RL gym for tool-driven editing and rubric-based verification. Reports that frontier MLLMs perform under 60% and that RL fine-tuning improves precise 3D editing.
A Gated-DeltaNet-aware mixed-precision GGUF quantization of Qwen3.8-27B for efficient local inference; preserves the MTP draft head and offers an optional BF16 mmproj for images. Weights are ~11.73 GiB (3.69 bpw), sized for 16–24 GB GPUs at modest context.
Trains a foundation GUI agent using a closed-loop, environment-grounded data stack plus in-context multimodal demonstrations to automate long-horizon desktop workflows. Combines scalable task generation/verification, subtask-level demo guidance, and a 100-task OSWorkerBench benchmark to improve strict success and task progress.
Analyzes how on-policy distillation (OPD) transfers teacher LLM capabilities to student models across in-domain shifts, cross-domain transfer, and multi-teacher settings. Key findings: OPD conveys reasoning patterns rather than specific answers, same-origin teacher-student pairs generalize broadly, and multi-teacher combinations induce mixture-dependent tradeoffs.
Presents a unified black-box reinforcement learning framework to train and optimize agents running inside complex execution harnesses. Uses sandbox-parallel rollouts, a serving proxy that captures model calls and reconstructs multi-turn trajectories as prefix trees, and adapted GRPO/PPO optimizers to achieve stable, scalable RL across heterogeneous harnesses.
Injects proprietary news, regulatory and legal data into an open checkpoint via data-centric continual learning to improve performance on legal, tax and journalism tasks while preserving general capabilities and very long context support.
Evaluates whether AI systems can independently carry out project-level scientific research by progressively removing human methodological guidance across 60 tasks in 11 domains. Built with expert review, sandbox execution, and multi-agent–model scoring to measure innovation and autonomous experimental execution.
A 9B open-weight reasoning LLM that uses a self-improvement loop to auto-generate tasks, construct scaffolds, and optimize rollouts for stronger agentic coding and long-context reasoning. Single-GPU deployable, supports tool-calling and a 262,144-token context window.
Turns embodied navigation into 2D visual prompting where a vision-language model selects image pixels that are projected to 3D actions; adds selective chain-of-thought, compressed anchor-trajectory memory, and a two-level alignment objective to improve sample and runtime efficiency.
Fine-tunes long-horizon LLM agents with evolution strategies so full-model updates run at inference-level GPU memory. Emphasizes trajectory-level credit via black-box rewards, online prompt–parameter co-evolution, and a cosine decay for perturbation scale to balance exploration and adaptation; suited for limited-GPU settings.
A 35B mixture-of-experts LLM tuned for agentic coding and end-to-end self-improvement: it jointly generates tasks, scaffolds, and solution rollouts. Activates ~3B params/token, supports 256K context (extendable), and emits chain-of-thought plus OpenAI-style tool calls.