Results for “pre-pr”

Simple Preference Optimization for LLM alignment. Reference-free alternative to DPO with better performance (+6.4 points on AlpacaEval 2.0). No reference model needed, more efficient than DPO. Use for preference alignment when want simpler, faster training than DPO/PPO.

Simple Preference Optimization for LLM alignment. Reference-free alternative to DPO with better performance (+6.4 points on AlpacaEval 2.0). No reference model needed, more efficient than DPO. Use for preference alignment when want simpler, faster training than DPO/PPO.

Mathematical AI interpretability with sheaf cohomology, spectral analysis, causal inference, and hallucination prevention

Prepare release readiness notes, version checks, test checklist, and local issue drafts.

Process large ML datasets in parallel across CPU or GPU clusters, with streaming execution, multi-format I/O, and integration with Ray Train, PyTorch, and TensorFlow for batch inference and preprocessing pipelines.

Predicts tissue or cell-type gene expression (log TPM and TPM) from a TSS-centered DNA sequence using the hosted Genomic Intelligence G0 Expression model, conditioned on a free-text cell-type description.

Detect splice donor and acceptor sites in DNA sequences using the Genomic Intelligence G0 BigBird transformer, via the hosted /v1/tasks/splice/predict API. Returns per-position site probabilities and called sites.

Audits ECC Tools GitHub App for cost issues like runaway PR creation, quota bypass, premium-model leakage, and duplicate jobs, using an evidence-first workflow.

Automatically rig 3D models with UniRig (VAST-AI-Research, SIGGRAPH'25) — predict a skeleton, predict skinning weights, and merge the rig back onto the original mesh. Use when the user wants auto-rigging for .obj/.fbx/.glb/.gltf/.dae/.vrm assets, a skeleton or skin weights for a character or creature, a UniRig environment prepared on a CUDA machine, batch rigging of a model directory, or an honest comparison between UniRig, SkinTokens, Tripo, Mixamo, AccuRig, and Blender Rigify. Triggers on: unirig, auto rig, auto-rigging, 3D rigging, skeleton prediction, skinning weights, rig a character, armature generation, rigged glb, rigged fbx, bone weights.

Build and maintain digital twins - virtual representations of physical systems that synchronize with real-world counterparts for monitoring, prediction, and optimization. Use when "digital twin, virtual model, real-time synchronization, physical-virtual coupling, predictive maintenance, asset modeling, system replica, live simulation, " mentioned.

Prepare and classify local test commands before asking approval to execute them.

Expert en montage vidéo (Premiere, DaVinci, CapCut, color grading, transitions, formats d'export)

Defines an inter-agent communication protocol for C-suite agent teams, including invocation syntax, loop prevention, isolation rules, and response formats to coordinate cross-functional analysis and board meetings.

Fast structured generation and serving for LLMs with RadixAttention prefix caching. Use for JSON/regex outputs, constrained decoding, agentic workflows with tool calls, or when you need 5× faster inference than vLLM with prefix sharing. Powers 300,000+ GPUs at xAI, AMD, NVIDIA, and LinkedIn.

Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.

Fast structured generation and serving for LLMs with RadixAttention prefix caching. Use for JSON/regex outputs, constrained decoding, agentic workflows with tool calls, or when you need 5× faster inference than vLLM with prefix sharing. Powers 300,000+ GPUs at xAI, AMD, NVIDIA, and LinkedIn.

Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.

Fast structured generation and serving for LLMs with RadixAttention prefix caching. Use for JSON/regex outputs, constrained decoding, agentic workflows with tool calls, or when you need 5× faster inference than vLLM with prefix sharing. Powers 300,000+ GPUs at xAI, AMD, NVIDIA, and LinkedIn.

Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.

Forecast any univariate time series (sales, sensors, energy, vitals, weather) zero-shot using Google's TimesFM foundation model, with point forecasts and prediction intervals from CSV, DataFrame, or array inputs.

编写评估FastText文本分类模型的Python函数，计算accuracy、F1、recall和precision指标，并处理特定格式的标签文本分割。

Run Karpathy-style autonomous ML search on a real training repo: choose the right mode (setup, program.md, bounded loop, results interpretation, or constrained-hardware adaptation), preserve the immutable prepare.py / 300-second / val_bpb contract, and route prompt/skill eval work away to LangSmith, Promptfoo, Braintrust, or skill-autoresearch.

Run cloud-native molecular modeling and drug-design workflows via a Python API, covering pKa prediction, docking, conformer and tautomer ensembles, molecular dynamics, and related small-molecule or protein tasks without local HPC infrastructure.

Designs and governs recursive self-improvement loops where an agent mines its own failures and proposes edits to its own harness, prompts, or workflow, covering acceptance gates, diversity preservation, and the optimization ladder.

Tail, read, or search a Stata log file from a previous command or background task.

Nonlinear dimensionality reduction with UMAP — fast manifold learning for 2D/3D visualization, clustering preprocessing (e.g., HDBSCAN), and supervised or parametric UMAP. Use when projecting high-dimensional data to low dimensions for visualization, embedding generation, or as a preprocessing step before clustering. Part of the AlterLab Academic Skills suite.

Diagnose and fix macOS resource pressure for Claude Code workflows. Use when load avg is high, swap is saturated, CC feels slow, or before spinning up parallel agents/worktrees. Covers CPU top-talkers, swap pressure, zombie claude processes, Node heap tuning, Spotlight/background-agent pruning, and purge. Apple Silicon aware.

Adds PyTorch FSDP2 (fully_shard) to training scripts with correct init, sharding, mixed precision/offload config, and distributed checkpointing. Use when models exceed single-GPU memory or when you need DTensor-based sharding with DeviceMesh.

Monitor a live model for data quality, input and prediction drift, performance decay, and fire retraining triggers.

Use the model-recommender skill, Workflow D (Setup Questionnaire), to configure model access and project-level preferences.

Provides guidance for setting up an Elasticsearch MCP server, including prerequisites and links to official documentation.

Fashion Tech Specialist IA — Expert en technologies mode (virtual try-on, size recommendation, trend prediction, sustainable fashion)

Pre-ingestion verification for epistemic quality in RAG systems with 9-point verification and Two-Round HITL workflow

Featurizes molecules for machine learning with molfeat (100+ featurizers) — ECFP/MACCS/MAP4 fingerprints, RDKit and Mordred physicochemical descriptors, and pretrained embeddings (ChemBERTa, ChemGPT, GIN) exposed as scikit-learn transformers that convert SMILES into feature vectors. Use when turning molecules into ML-ready feature matrices for QSAR/QSPR or virtual screening, or benchmarking fingerprint against descriptor and embedding representations; for training models and MoleculeNet benchmarks on those features prefer alterlab-deepchem, and for low-level fingerprint or descriptor primitives prefer alterlab-rdkit. Part of the AlterLab Academic Skills suite.

Compute the F1Score metric using torchmetrics when predictions and ground-truth labels are available.

Design, implement, harden, and verify Model Context Protocol (MCP) servers with precise tool contracts, least-privilege authorization, safe transports, structured errors, and interoperability tests. Use when creating a new MCP server, exposing an API or data source through MCP, reviewing an MCP server design, adding or revising MCP tools, or preparing an MCP server for production.