Models from the Hydra paper: dual-head VLM combining retrieval + generation. Main, variants, baselines, ablations.
Athrael Soju PRO
athrael-soju
AI & ML interests
Yes
Recent Activity
liked a dataset about 16 hours ago
Jackrong/Qwen3.5-reasoning-700x liked a dataset about 16 hours ago
nohurry/Opus-4.6-Reasoning-3000x-filtered updated a Space about 19 hours ago
athrael-soju/HydraQwen3.5-0.8B-demoOrganizations
None yet
ColBERT-style late-interaction visual document retrieval adapters built on Google Gemma-4 (E2B and E4B variants).
Visual document retrieval models on Qwen3.5 backbone. ViDoRe v3 leaderboard competitors, 128-dim multi-vector.
-
Spatially-Grounded Document Retrieval via Patch-to-Region Relevance Propagation
Paper • 2512.02660 • Published -
Architecture-Aware LLM Inference Optimization on AMD Instinct GPUs: A Comprehensive Benchmark and Deployment Study
Paper • 2603.10031 • Published -
Hydra: Unifying Document Retrieval and Generation in a Single Vision-Language Model
Paper • 2603.28554 • Published
Models from the Hydra paper: dual-head VLM combining retrieval + generation. Main, variants, baselines, ablations.
Visual document retrieval models on Qwen3.5 backbone. ViDoRe v3 leaderboard competitors, 128-dim multi-vector.
ColBERT-style late-interaction visual document retrieval adapters built on Google Gemma-4 (E2B and E4B variants).
-
Spatially-Grounded Document Retrieval via Patch-to-Region Relevance Propagation
Paper • 2512.02660 • Published -
Architecture-Aware LLM Inference Optimization on AMD Instinct GPUs: A Comprehensive Benchmark and Deployment Study
Paper • 2603.10031 • Published -
Hydra: Unifying Document Retrieval and Generation in a Single Vision-Language Model
Paper • 2603.28554 • Published