ADu2021/skillXiv - 16페이지

llm-finetuning-dataset-synthesis

Automatically generate high-quality domain-specific fine-tuning datasets from raw documents using adaptive processing and persona-driven synthesis, preserving general capabilities while specializing models.

원문 언어: 영어

llm-forecasting-evaluation

Benchmark LLMs on real-world forecasting questions from Metaculus, comparing against human crowds and expert forecasters. Identifies which domains LLMs handle well and where they fall short relative to human intelligence.

원문 언어: 영어

llm-local-linear-mappings

Interpret LLM behavior as locally linear mappings between hidden representations, enabling mechanistic understanding of computation without examining individual weights or attention patterns.

원문 언어: 영어

llm-memorization-landscape

Understand and mitigate unintended memorization of training data in LLMs by systematizing mechanisms, detection methods, and mitigation strategies across the model lifecycle.

원문 언어: 영어

llm-sandbox-agentic-intelligence

Enable language models to explore code sandboxes to solve diverse tasks like mathematics and long-context reasoning without additional training, with reinforcement learning further enhancing capabilities. Use when you need LLMs to perform tasks requiring…

원문 언어: 영어

llm2vec-gen-response-embeddings

Learn to encode LLM-generated responses rather than raw inputs by training special tokens and projection layers while keeping the backbone frozen. Bridges the input-output gap to transfer LLM capabilities like reasoning directly into embedding space.

원문 언어: 영어

locality-aware-parallel-autoregressive-decoding

Accelerate image generation by generating multiple patches in parallel instead of sequentially. Uses locality-aware ordering that generates spatially-close tokens while keeping concurrent tokens far apart, reducing steps from 256 to 20 (12× speedup) while…

원문 언어: 영어

locate-steer-and-improve-a-practical-survey-of

Mechanistic Interpretability (MI) has emerged as a vital approach to demystify the opaque decision-making of Large Language Models (LLMs). However, existing reviews primarily treat MI as an observational science, summarizing analytical insights while lacking…

원문 언어: 영어

log-linear-sparse-attention

Reduce self-attention complexity from O(N²) to O(N log N) through hierarchical token selection and enrichment. Perform hierarchical Top-K selection progressively adopting sparse Top-K at each level. Implement sparse index transpose algorithm avoiding dense…

원문 언어: 영어

long-context-rl-reasoning

Train LLMs for long-context reasoning using KeyChain synthesis: convert short multi-hop QA into long-context tasks by embedding UUID chains in distractor documents, enabling 16K→128K generalization.

원문 언어: 영어

long-cot-training

Navigate Long CoT Degradation phenomenon when training small models on extended reasoning, understanding recovery dynamics and implementing strategies to maintain performance.

원문 언어: 영어

long-term-agent-memory

Implements Controllable Memory Usage from arXiv:2601.05107

원문 언어: 영어

longanimation-dynamic-memory-generation

Generate animations longer than 500 frames with consistent coloring. Uses dynamic global-local memory to compress long-term history and intelligently fuse global color features with local generation, enabling 49% quality improvement over previous methods.

원문 언어: 영어

longcat-flash-thinking-2601-technical-report

Implement techniques from LongCat-Flash-Thinking-2601 Technical Report. LongCat-Flash-Thinking-2601 achieves state-of-the-art performance among open-source models on a wide range of agentic benchmarks, including agentic search, agentic tool use, and…

원문 언어: 영어

longcontext-reward-modeling

Train reward models to evaluate long-context responses by introducing faithfulness as a criterion and using consistency-based alignment to maintain judgment-explanation coherence.

원문 언어: 영어

longllada-diffusion-context

Training-free method extending diffusion LLMs to 6x context length using NTK-based RoPE scaling, exploiting bidirectional attention stability.

원문 언어: 영어

longvie-multimodal-video-generation

Generate consistent ultra-long videos (up to one minute) using depth and trajectory controls with autoregressive segment generation and unified noise initialization.

원문 언어: 영어

longvila-scaling-rl-long-videos

Train vision-language models on hour-long videos using a two-stage pipeline combining supervised fine-tuning with GRPO, reaching 71% accuracy on VideoMME while supporting 8,192 frames through efficient multi-modal sequence parallelism.

원문 언어: 영어

lookaheadkv-predictive-cache-eviction

Evict unnecessary KV cache entries using trainable lookahead tokens and LoRA modules that predict token importance without drafting responses. Achieve 14.5× speedup over draft-based methods with 2% eviction overhead.

원문 언어: 영어

looped-elastic-depth-transformers

Enable budget-conditioned reasoning by repeatedly applying a shared transformer block stack with trajectory-based conditioning on time and step size. Train via shortcut-consistency loss to align shorter and full-length trajectories, enabling variable-depth…

원문 언어: 영어

looptool-closed-loop-tool-learning

Improve LLM tool-use through automated, closed-loop data curation combining capability probing, error correction, and targeted data expansion—enabling self-refining training pipelines that surpass larger models without expensive APIs.

원문 언어: 영어

lopa-lookahead

Maximize parallelism in diffusion language models by intelligently exploring token filling orders. Spawn multiple candidate branches predicting high-confidence positions, evaluate all branches in one forward pass, and select the branch enabling most future…

원문 언어: 영어

low-prob-exploration

Preserve reasoning diversity in RL by protecting low-probability tokens that represent novel thinking paths. Trigger: maintain exploration capability during RL training while avoiding entropy collapse on reasoning tasks.

원문 언어: 영어

lspo-length-aware-dynamic-sampling

Filter training samples by response length to identify high-confidence correct solutions and complex problems, improving sample efficiency in LLM reasoning RL without additional metrics.

원문 언어: 영어

lsrif-logic-structured-rl

Improves instruction-following by using differentiated reward mechanisms that recognize logical dependencies (sequential, conditional, parallel) in instructions, enabling better model reasoning about task structure.

원문 언어: 영어

ltx-2-efficient-joint-audio-visual-foundation-mode

Multimodal system for joint audio-video understanding and generation, enhancing agent perception and communication in environments with both audio and visual information.

원문 언어: 영어

lumine-generalist-agents-3d-worlds

Build generalist agents for 3D open worlds using vision-language models with human-like interaction patterns and adaptive reasoning—achieving zero-shot transfer across games without fine-tuning through pixel-level perception and high-frequency action…

원문 언어: 영어

lumos-1-autoregressive-video

Generate videos autoregressively by extending LLM architectures to spatiotemporal data. MM-RoPE balances frequency spectra across temporal and spatial dimensions, while Autoregressive Diffusion Forcing enables efficient parallel decoding. Lumos-1 (0.5B-3B…

원문 언어: 영어

luspo-length-unbiased-sequences

Fix length bias in sequence-level policy optimization by scaling each sequence's loss by its token count, eliminating gradient imbalances that cause models to shorten responses during training.

원문 언어: 영어

lychee-decode-sparse-kv-sharing

Classify attention heads into retrieval (full attention) and sparse (token-selected) roles using HardKuma distribution for differentiable discrete optimization. Sparse heads reuse KV pairs from retrieval heads, reducing cache by 90% while maintaining quality…

원문 언어: 영어

lynx-dynamic-exits

Enable models to stop generating when confident through lightweight hidden-state probes with distributional guarantees. LYNX achieves cross-domain transferability without retraining—ideal when you need confidence-controlled reasoning efficiency.

원문 언어: 영어

m3-bench-tool-use

Design and evaluate multimodal agents for tool use with M3-Bench: assess three interconnected dimensions (multi-modal grounding, multi-hop causality, multi-threaded parallelism) using similarity-bucketed Hungarian alignment for transparent tool call…

원문 언어: 영어

m4olgen-molecular-generation

Generates molecules meeting precise numeric property constraints across multiple dimensions through two-stage multi-agent framework with fragment-level edits and Group Relative Policy Optimization, improving validity and property satisfaction.

원문 언어: 영어

mact-document-understanding

Decomposes document processing into specialized agents (planning, execution, judgment, answer) with agent-wise adaptive test-time scaling. Achieves 9.9-11.5% performance gain with smaller models while maintaining reasoning.

원문 언어: 영어

magistral-reasoning-rl

Build reasoning capabilities through pure RL without distilled traces, achieving 50% AIME accuracy improvement via scalable asynchronous training with novel reward shaping for multilingual consistency.

원문 언어: 영어

magma-a-multi-graph-based-agentic-memory-architect

Memory management architecture using multi-graph representations for autonomous agents, enabling efficient knowledge organization, contextual retrieval, and dynamic memory expansion to support complex agent decision-making.

원문 언어: 영어

mai-ui-agents

Scale GUI agents to real-world complexity via extended action space (user interaction, tool calls) and device-cloud collaboration. Online RL supports 500+ parallel environments with asynchronous handling; local agent monitors trajectory alignment and handoffs…

원문 언어: 영어

mano-manifold-optimization

Improve LLM training efficiency through manifold-based optimization that projects momentum onto tangent spaces and constrains updates on rotational Oblique manifolds. Achieves 1.75× faster convergence than Muon with reduced memory.

원문 언어: 영어

mapo-mixed-advantage-policy-optimization

Dynamically reweight advantage functions based on trajectory certainty to improve policy optimization in foundation models. Addresses advantage reversion and mirror problems by mixing standardized and mean-normalized advantage formulations. Enables more…

원문 언어: 영어

margin-adaptive-dpo

Adapt DPO temperature parameters per sample using reward model margins, amplifying learning signals for hard preference pairs while dampening easy ones.

원문 언어: 영어