Uses Parallel Looped Transformers (PLT) to make loop count a practical knob for code models, finding two loops give the best test-time gains. Trains 7B models on 18T tokens and attributes saturation beyond two loops to a gain–cost tradeoff from positional mismatch.
Assesses whether coding agents can generate complete, playable games end-to-end inside the Godot engine. Implements an interaction-grounded evaluation (replayed demonstrations + rubric-guided multimodal judging) across 140 tasks and 15 game families; top agents score ~41%.
Contains 603 coding and math prompt–response pairs produced by Claude Fable‑5 (generated 2026-06-10), provided as a JSONL subset for fine-tuning, evaluation, and behavior analysis. Responses are 'non-thinking' (no chain-of-thought); small, anonymized, and lacking an explicit license.
Serves interactive, long-lived streaming video-generation sessions by jointly scheduling session placement and GPU autoscaling to meet tight per-chunk latency. Combines migration-aware placement, load-driven autoscaling, coalesced chunk processing, GPU–CPU offloading and NCCL GPU–GPU migration; reports ~37% reductions in worst-case per-chunk latency and GPU operating cost.
A local, quantized Gemma 4 12B fine-tune packaged as GGUF quants that runs on ~4.5 GB VRAM. Optimized for coding and multi-step agentic tool use (read→reason→act→verify), ships multiple quant sizes (Q3_K_M–Q8_0) and MTP draft support; English-centric with trade-offs versus generalist models.
A 35B mixture-of-experts LLM specialized for agentic coding and tool-enabled code generation, fine-tuned with self-scaffolding reinforcement learning. Supports very long contexts, OpenAI-compatible tool calls, and multiple serving runtimes under an MIT license.
Compresses an entire document collection into 16 query-dependent memory tokens for a frozen LLM. It retrieves useful clues across long histories while keeping the model’s input length fixed.
Compares 30 frontier LLMs generating static SVG markup from 500 prompts using 1,355,161 human votes across three leaderboards (Preference, Coherence, Alignment); provides raw SVGs, 768×768 rasterized PNGs, and per-comparison human vote records under a CC-BY-4.0 prompt license.
Provides a GGUF-quantized local build of Ornith-1.0's 9B dense model for offline inference and terminal-focused coding agents. Supports OpenAI-compatible tool-calling, a 256K context window, and runs via llama.cpp or Ollama on a single high-memory GPU.
Provides a rubric-based benchmark that converts dense image captions into instance-specific atomic checks (Must-Right and Easy-Wrong) and a gated scoring rule, aiming to expose perceptual brittleness and better align multimodal model evaluation with human judgment.
Provides executable programs for generating 3D content, paired with captions, renders, and available verification records. It spans multiple 3D coding environments, with source-specific licenses and varying sample quality.
Provides runnable 3D programs with renders, descriptions, and metadata for training and evaluating models that generate 3D code from text or images. Dataset access is restricted to non-commercial research.