When Losses Align: Gradient-Based Composite Loss Weighting for Efficient Pretraining
Paper • 2605.07756 • Published • 1
Joel Wang
joelhenwang
AI & ML interests
None yet
Recent Activity
upvoted a paper about 5 hours ago
When Losses Align: Gradient-Based Composite Loss Weighting for Efficient Pretraining upvoted a paper 1 day ago
RoPE-Aware Bit Allocation for KV-Cache Quantization upvoted an article 2 days ago
Kog Laneformer 2B: The Latency-First Model Behind Kog Inference EngineOrganizations
upvoted a paper about 5 hours ago
upvoted a paper 1 day ago
upvoted an article 2 days ago
Article
Kog Laneformer 2B: The Latency-First Model Behind Kog Inference Engine
kogai
• • 27 upvoted 6
papers 4 days ago
Q-Delta: Beyond Key-Value Associative State Evolution
Paper • 2606.08804 • Published • 1
Comparing Linear Probes with Mahalanobis Cosine Similarity
Paper • 2606.19603 • Published • 3
Demystifying Training-Time Augmentation for Data-Constrained Language Model Pretraining
Paper • 2606.16246 • Published • 4
A Verifiable Search Is Not a Learnable Chain-of-Thought
Paper • 2606.21884 • Published • 3
Tmax: A simple recipe for terminal agents
Paper • 2606.23321 • Published • 12
Learning from Your Own Mistakes: Constructing Learnable Micro-Reflective Trajectories for Self-Distillation
Paper • 2606.18844 • Published • 18
upvoted 11
papers 5 days ago
TANDEM: Bi-Level Data Mixture Optimization with Twin Networks
Paper • 2606.04401 • Published • 1
Always Learning, Always Mixing: Efficient and Simple Data Mixing All The Time
Paper • 2605.15220 • Published • 1
Towards Efficient LLMs Annealing with Principled Sample Selection
Paper • 2605.31175 • Published • 1
KaLM-Reranker-V1: Fast but Not Late Interaction for Compressed Document Reranking
Paper • 2606.22807 • Published • 47
When, Where, and How: Adaptive Binning for Tabular Self-Supervised Learning
Paper • 2606.19827 • Published • 3
FastMix: Fast Data Mixture Optimization via Gradient Descent
Paper • 2606.14971 • Published • 3
Tapered Language Models
Paper • 2606.23670 • Published • 8
Causal Discovery in the Era of Agents
Paper • 2606.23608 • Published • 7
Connect the Dots: Training LLMs for Long-Lifecycle Agents with Cross-Domain Generalization Via Reinforcement Learning
Paper • 2606.20002 • Published • 8
Training Open Models for Agentic Phone Use
Paper • 2606.23049 • Published • 15
Self-Compacting Language Model Agents
Paper • 2606.23525 • Published • 17