Discover the Best AI Resources
Curated essentials, no noise — just what matters
Compresses a 27B-class multimodal model into end-to-end ternary weights to run 27B reasoning on-device: 5.9–8.6 GB deployed footprint, 262K-token context, ~98.2% of FP16 benchmark performance; ships MLX and GGUF packs and runs on Apple MLX and CUDA.
Turns Qwen3.5 into a 3-way NLI cross-encoder (entailment/contradiction/neutral) for zero-shot reranking, grading, content guarding, or action selection. Provides a pretrained Qwen3.5-4B checkpoint, helper utilities (OpenJevCrossEncoder, LatentMLPHead), and optional 35B MoE latent heads for per-task scoring.
Evaluates whether MiniMax-H3 can infer latent physical events from incomplete multimodal evidence and produce videos consistent with those inferences. Builds a 517-instance benchmark across four reasoning scenarios (multi-view, audio disambiguation, video continuation, audiovisual integration) and quantifies modality-specific failure modes.
Benchmark for typed probabilistic decisions: given one shared state, a model answers multiple typed questions at once and returns full probability distributions (noul/choice/score). Contains four workflows with train/test splits and soft gold labels from teacher samples, designed to evaluate accuracy, calibration, and latency trade-offs.
Runs locally on constrained devices to turn text into guaranteed-parsable JSON tool calls, typed structured extractions, or sentence embeddings. Delivered as a single compact weights file (8–29 MB) with a laddered 2–20-layer design, low-bit quantisation and calibrated confidence scores for on-device apps.
A 27B-class language model packaged in GGUF with end-to-end ternary weights for on-device or single-GPU llama.cpp inference; reduces FP16 footprint to ~5.9–7.2 GB while retaining ~98% of baseline performance and supporting up to 262K tokens.
A 29B-parameter mixture-of-experts LLM for long-context reasoning and agent-style workflows, with native 256K context and Ascend NPU co-optimizations. Compatible with Transformers/vLLM ecosystems; suitable for long-horizon agents and domain fine-tuning.
Provides transcripts, translations, and multi-task text labels for speech across 48 languages and about 463,000 hours. The release contains manifests only; audio must be obtained from the original data providers.
Provides a programming model and distributed runtime that preserves parent–child lineage and ordered variable-cardinality expansions for 1→M→1 dataflow pipelines, enabling completion-driven cross-input GPU batching and ordered gathers for foundation-model data preparation. Demonstrates multi-GPU scaling speedups and lower end-to-end time versus Ray Data and Daft.
Evaluates multi-field JSON schemas in parallel to extract boolean or categorical field values from text, producing guaranteed-valid JSON and per-field calibrated confidences. Uses KV-cache broadcasting, sub-vocabulary logit slicing and token-tree disambiguation to cut latency (5.6x–7.0x on M4 Max) versus autoregressive decoding; requires Apple Silicon and MLX.
Introduces physical rank consistency (PRC) and ActionPiece, a learned action tokenizer that preserves local physical relationships via joint supervision of representation learning and quantization, improving autoregressive vision-language-action policy performance across multiple benchmarks.
Transforms scientific code repositories into executable, agent-learnable environments that support task generation, execution, and scientific verification. Agent-guided repository transformation produces verified interaction trajectories used to train the PhAI-IDE model family. Intended for research on agent learning, scientific-code repair, and training RL/SFT models.