🧠 LLM Training - inarcissuss

Transformer-Based Language Models Across Domain Verticals: Architectures, Applications and Critical Assessment

🧠LLM Research arXiv·

Jolia: Concept-Level Vision-Language Alignment for 3D CT Contrastive Learning

🧠LLM Research arXiv·

PoLAR: Factorizing Extent and Mode in Latent Actions for Robot Policy Learning

Covered by Turing Post

🧠LLM Research arXiv·

Characterizing Narrative Content in Web-scale LLM Pretraining Data

🎯RLHF arXiv·

Self-Evolution for Multi-Turn Tool-Calling Agents via Divergence-Point Preference Learning

🧠LLM Research arXiv·

Comparing Transformers and Hybrid Models at the Token Level

Covered by Hugging Face, Ai2

🧠LLM Research arXiv·

HumanScale: Egocentric Human Video Can Outperform Real-Robot Data for Embodied Pretraining

Covered by topicqueue.substack.com

🔬ML Research arXiv·

Is Our Benchmark Enough? An Analysis of Continual Learning for MLLMs

🧠LLM Engineering arXiv·

Techniques for Peak Memory Reduction for LoRA Fine-tuning of LLMs on Edge Devices

🗣️Large Language Models arXiv·

Can LLMs Reliably Self-Report Adversarial Prefills, and How?

🧠LLM Research arXiv·

Open Problem: Is AdamW Effective Under Heavy-Tailed Noise?

🎯Alignment Research arXiv·

Uncertainty-Aware Reward Modeling for Stable RLHF

🧠Transformer Architecture arXiv·

Parallel Manifold Steering: Efficient Adaptation of Large Associative Memories via Residual Energy Shaping

🎯RLHF arXiv·

Beyond Uniform Forgetting: A Study of Sequential Direct Preference Optimization Across Preference Settings

🗣️Large Language Models arXiv·

Skin-Deep: A Geometric Diagnostic for Alignment Fragility in Large Language Model Representations

🧠LLM Research arXiv·

LLM-Guided Test-Time Discovery of Quantum-Chemical Approximation Algorithms

🧠LLM Research arXiv·

IHUBERT: Vector-Based Semantic Deduplication and Domain-Balanced Pretraining for Persian Resources

🗣️Large Language Models arXiv·

Latent Personal Memory: Represent personal memory as dynamic soft prompts

🧠LLM Research arXiv·

Yuvion VL: A Multimodal Foundation Model for Adversarial Content and AI Safety

Transformer-Based Language Models Across Domain Verticals: Architectures, Applications and Critical Assessment

Jolia: Concept-Level Vision-Language Alignment for 3D CT Contrastive Learning

PoLAR: Factorizing Extent and Mode in Latent Actions for Robot Policy Learning

Characterizing Narrative Content in Web-scale LLM Pretraining Data

Self-Evolution for Multi-Turn Tool-Calling Agents via Divergence-Point Preference Learning

Comparing Transformers and Hybrid Models at the Token Level

HumanScale: Egocentric Human Video Can Outperform Real-Robot Data for Embodied Pretraining

Is Our Benchmark Enough? An Analysis of Continual Learning for MLLMs

Techniques for Peak Memory Reduction for LoRA Fine-tuning of LLMs on Edge Devices

Can LLMs Reliably Self-Report Adversarial Prefills, and How?

Open Problem: Is AdamW Effective Under Heavy-Tailed Noise?

Uncertainty-Aware Reward Modeling for Stable RLHF

Parallel Manifold Steering: Efficient Adaptation of Large Associative Memories via Residual Energy Shaping

Beyond Uniform Forgetting: A Study of Sequential Direct Preference Optimization Across Preference Settings

Skin-Deep: A Geometric Diagnostic for Alignment Fragility in Large Language Model Representations

LLM-Guided Test-Time Discovery of Quantum-Chemical Approximation Algorithms

IHUBERT: Vector-Based Semantic Deduplication and Domain-Balanced Pretraining for Persian Resources

Latent Personal Memory: Represent personal memory as dynamic soft prompts

Towards Engineering Scaling Laws with Pretraining Data Composition