Claude Skills by ADu2021

Represent 3D scenes as compressed light-field tokens for efficient neural rendering. Multi-view images are tokenized via Plücker coordinates, condensed through K-means clustering, and rendered adaptively. Achieves 5-7× data reduction versus MVSplat while enabling on-the-fly quality-speed tradeoffs: up to 66% FPS improvement with controlled token counts.

Augment verifiable reward RL (RLVR) with contrastive learning to generate dense auxiliary rewards. Enforce proximity among correct reasoning trajectories in embedding space while suppressing errors, amplifying invariant reasoning patterns.

Replace hard clipping in policy gradients with smooth quadratic penalties derived from Total Variation divergence constraints. Eliminates zero-gradient regions and training instability while maintaining stable policy evolution.

No More Stale Feedback: Co-Evolving Critics for Open-World Agent Learning. From arXiv:2601.06794

Hybrid multi-agent architecture where orchestrator delegates tasks to GUI Operator or Programmer agent. Coding enables efficiency on computational tasks, achieving 60.76% on OSWorld with 33% fewer steps.

Dynamically allocate training budget across samples using a capability-oriented value function that measures per-sample training importance based on model capability evolution. Reduces training time via greedy heap-based allocation optimizing exploration-exploitation tradeoff.

Generates complex text-heavy and structured images by converting natural language descriptions into executable code that specifies layouts, then renders and refines. Achieves 68% improvement on structured image generation benchmarks.

Train dual-brain agents with specialized planner (Cerebrum) and executor (Cerebellum) through decoupled RL, resolving planning-execution trade-off for scientific GUI agents

Train code and test generators through adversarial co-evolution where test LLM generates adversarial test cases to expose code defects. Prevent self-collusion by separating models and enabling white-box test generation.

Predict GUI state evolution by generating HTML code rather than pixel images. Combines visual fidelity of pixel-based approaches with structural precision of code-based methods through deterministic rendering. Enables agents to evaluate action consequences and select best decisions before execution.

Generate physically grounded 4D scenes from natural language through dual-stream architecture separating object detail from scene orchestration. VLM-Motion Critic validates physics parameters iteratively, bridging semantic-physical execution gap.

Generate Verilog hardware code from natural language using reasoning-enhanced LLMs, combining rule-based testbench generation with round-trip data synthesis and adaptive DAPO reinforcement learning for reliable hardware design.

Evaluate LLM coding capabilities across three dimensions—problem analysis, code implementation, and test validation—to identify specific reasoning gaps and improve model robustness through ensemble approaches.

Advanced reasoning approach for optimizing inference efficiency through meta-cognitive planning, enabling agents to make better decisions with reduced computational overhead.

Align VLA efficiency with human cognition through 3-stage progressive routing: instruction-aware aggregation, instruction-irrelevant pruning, and coupled attention for 2.8x inference speedup

Dynamically adapt LLM depth per input at test time by skipping, repeating, or reordering layers using MCTS search, correcting 60% of initially wrong predictions and processing 75% of correct predictions with shorter architectures without retraining.

Iteratively retrieve and reason over long narratives using a dynamic memory workspace that integrates retrieved facts into a shared context for complex multi-hop reasoning.

Organize agent memory as an event graph with explicit logical relationships rather than flat embeddings. Framework incrementally segments experiences into events and links them through causal, temporal, and logical relations. Enables agents to navigate memory as a logic map for goal-directed searching and structured reasoning, improving performance on multi-hop reasoning and long-horizon planning tasks.

Build generalist judge models for evaluating LLM outputs using verifiable rewards and policy gradient training. Create a 7B model competitive with much larger judges through reward-guided optimization and critical thinking decomposition. Use when you need reliable automated evaluation of model outputs across diverse tasks.

Lightweight verifier model for evaluating LLM outputs across multiple domains, handling diverse answer types through meta-error pattern analysis.

Improve RL sample efficiency through co-evolution of policy and experience extractor, enabling dynamic experience replay that adapts to the agent's skill level.

Decompose complex image editing instructions into simpler sub-tasks with automatically generated control guidance. Handles multi-object edits, preserves identity of surrounding regions, and eliminates manual mask creation.

Convert easy, high-accuracy training prompts into harder compositional problems by sequentially chaining multiple prompts together. Use Composition-RL to maintain effective learning signals during RL training when many prompts achieve near-perfect accuracy, enabling curriculum learning through progressive compositional depths.

Enable computer-use agents to flexibly choose between GUI primitives (click, type) and high-level tool calls, reducing cascading errors by 22% and improving execution speed by 11%.

Enable AI agents to safely explore action outcomes before execution by predicting UI state changes in desktop applications. Two-stage approach: first predict textual description of what changes, then synthesize visual representation of resulting screen. Allows agents to compare multiple candidate actions without risky trial-and-error, trained on Microsoft Office interactions (Word, Excel, PowerPoint).

Learns richer spatial representations by training on both 2D and 3D data simultaneously. Combines 3D intra-modal self-distillation with 2D-3D cross-modal joint embedding, achieving 14.2% and 4.8% improvements over single-modality baselines in scene understanding and geometric consistency.

Improve language model reasoning using only model confidence as reward signals, eliminating need for labels or preference models while achieving substantial gains with minimal data.

Learn optimal configurations for agentic AI systems through hierarchical RL that treats configuration as a query-wise decision problem. Structure policy selects workflows/tools/budgets while prompt policy composes specific instructions, achieving 25% accuracy improvement with 35% cost reduction.

Improve student model robustness under covariate shift by using diffusion-based augmentation that targets spurious features via teacher-student disagreement.

Improves speculative decoding acceptance rates by exposing target model's intermediate reasoning through contemplate tokens. Achieves 8-11% acceptance rate improvement over EAGLE through future-direction guidance without extra forward passes.

Build AI code agents that scale to massive repositories with long-context reasoning and persistent memory. Confucius SDK achieves 59% Resolve@1 on SWE-Bench-Pro—ideal when AI needs to handle real-world codebases with complex toolchains.

Train adaptive ODE solvers that generate fast diffusion previews while maintaining consistency with full-step refinement. Learn context-aware integration coefficients through PPO without distilling base models. Achieve high-quality previews in few steps with 47% fewer steps than standard methods.

Train a specialized continue-thinking token via reinforcement learning to extend reasoning depth during inference, achieving superior performance over fixed-token baselines.

Enable dynamic control over reasoning depth during inference using Reasoning Control Fields that guide long chain-of-thought generation based on execution and quality parameters.

Joint optimization of policy and reward models in LLM reinforcement learning by leveraging rule-based reward precision and dynamically constructing training pairs to prevent reward hacking and improve performance.

Enhance spatial reasoning in multimodal LLMs by integrating depth and segmentation as auxiliary modalities with adaptive reasoning strategies. COOPER achieves 6.91% improvement in spatial understanding—when you need 3D-aware vision-language capabilities.

Implement techniques from CooperBench: Why Coding Agents Cannot be Your Teammates Yet. Resolving team conflicts requires not only task-specific competence, but also social intelligence to find common ground and build consensus

Enhance reasoning models by integrating executable code within thinking traces, enabling grounded computation verification and reducing hallucination in mathematical and logical reasoning.

Achieve efficient neural networks via self-supervised dynamic routing using Cosine Incompatibility Ratio (CIR). Ground gating decisions in geometric novelty rather than learned heuristics, enable per-sample/per-block binary routing via Gumbel-softmax, constrain with progressive FLOPs regularization—maintaining accuracy while reducing computation 28.5% on CIFAR-10.

Enable open-source reasoning model development with a 100K-instance Long CoT Collection, scaling from 1K o1 seed samples through guided synthesis with GPT-4o, achieving 2-3× RL performance gains.

Enable vision-language models to perform embodied question answering in 3D environments through active camera exploration. CoV uses training-free test-time reasoning to iteratively select relevant viewpoints and adjust camera angles until sufficient context is gathered, achieving 11-13% accuracy improvements across spatial reasoning benchmarks.

CoVe synthesizes high-quality tool-use training data using explicit task constraints as both generation guidance and verification validators, enabling effective agent training without manual curation.

Enhance language model reasoning through coupled sampling from prior (question-only) and posterior (answer-conditioned) distributions. Construct composite distribution mixing both at token level using hybrid sampling. Combine reconstruction term, selective NLL loss, and KL regularization. Achieve 12.4% improvement over base model and 2.3% over comparable baselines.

Optimize approximate nearest neighbor search via contrastive RL, learning to generate efficient code for HNSW graph construction, search, and refinement.

Permanently remove unwanted concepts from LLMs by identifying and suppressing sparse autoencoder features across layers, creating parameter-level changes that prevent reversal.

Improve formal theorem proofs by treating criticism—evaluation of semantic correctness—as a learning signal. Train critic models to distinguish correct from incorrect formalizations, then use their feedback to guide RL-based proof generation.

Improve LLM reasoning by combining numerical and natural language critique feedback in online RL for policy refinement.

Trains language models to provide quality feedback through two-stage RL. Stage 1 optimizes discriminability (distinguishing good vs bad responses). Stage 2 adds helpfulness rewards (improving actor after feedback). Achieves 9.02% improvement without requiring stronger supervisors for training data.

Create a universal memory infrastructure enabling agents across different frameworks to share experience trajectories without retraining. Improve agent performance by retrieving workflows from related domains and applying diagnostic fixes.

Build desktop agents via reusable, parameterized skills encoding human computer-use knowledge. Skills combine execution graphs (handling UI variations) with composition graphs (chaining strategies). 57.5% success on WindowsAgentArena.