Turns solved protein structures into FoldingCorpus and Fold2Reason — a post-training recipe that supervises an LLM with discrete structural Q&A plus continuous 3D geometry to improve spatial, graph and scientific reasoning; reports consistent gains across 10 benchmarks.
Delivers fast, calibrated 'System 1' decisions over text and images (yes/no, 2–256 choices, 0–5 scores) in one forward pass, plus full 'System 2' Qwen3.8-27B multimodal reasoning. Highlights: zero-shot vision judging, 256K-token context, vLLM decision API, Apache-2.0.
Parses documents and answers questions about their contents with a 0.8B-parameter model. It combines OCR, structured extraction, and visual question answering, including support for scanned, skewed, and photographed pages.
Turns a structured state (text/JSON/images/video) plus a schema of typed questions into probabilistic decisions in one forward pass. Multimodal, Jev/SystemOne-compatible, 64k context window, post-trained from Qwen3.8-27B; no free-form text output.
Generates quantized point-and-box visual primitives for precise grounding across detection, referring, OCR, GUI and layout tasks; uses a shared coordinate vocabulary and staged pretraining to improve downstream robotic manipulation and autonomous driving when used as a visual backbone.
Uses a multimodal model's own critiques as privileged context and applies on-policy self-distillation over diffusion sampling trajectories to internalize corrective guidance, improving text-to-image generation without an external teacher; shows measurable gains on GenEval and GenEval2.
Analyzes how proposer–solver loops in self-evolving search agents can develop shared errors (co-cheating) that inflate internal rewards; introduces Multi-Sample Verification and CrossFit (cross-fitted scoring with partitioned sources) to reduce false agreement and improve downstream search performance.
Converts a state and a typed-question schema into probabilistic, structured decisions from text, JSON, images, or video—returning per-option probabilities for noul/choice/score questions in a single forward pass. Post-trained from Qwen3.5-9B and compatible with Jev/SystemOne APIs; optimized for low-latency decisioning.
Adaptive-granularity post-retrieval evidence compressor for multimodal RAG that selects variable-sized regions from retrieved text, tables, images, and videos to retain necessary context while cutting irrelevant content. Key features include a hierarchy-based node encoder with parent-relative refinement and a critic that issues targeted follow-up retrievals; yields higher QA accuracy across five benchmarks and reduces reader-input tokens by ~14–28%.
Performs a byte-level transplant of 144 tensors in an already-quantized GSQ-RCO Qwen3.8-Flash-Next to ablate the model's refusal direction while preserving GSQ-learned scales and the upstream per-tensor type assignment; multimodal, 262K context. Intended for local inference, red-teaming and quantization research; no retraining or built-in safety.
Post-trained multimodal Qwen3.8-27B variant that uses alternating SFT and RLOO to reduce pathological long reasoning tails; ships multiple quantization tiers (BF16, FP8, NVFP4, INT8, INT4, GGUF), supports MTP and DFlash2 speculative decoding, and includes detailed benchmark and runtime recommendations.
Provides a locally deployable, FP4-quantized variant of Qwen3.8-Flash-Next for multimodal (text, image, video) text generation. Uses NVIDIA Model Optimizer with quantization-aware distillation (QAD), a ~96 GiB checkpoint, and is validated on vLLM for Blackwell and DGX Spark.