Deep Fusion Memory adds chunk-level parametric memory to LLMs
With the backbone frozen, training needs 32 times fewer retrieval queries and inference uses no explicit retrieval.
With the backbone frozen, training needs 32 times fewer retrieval queries and inference uses no explicit retrieval.
Athanor agents autonomously post-trained Qwen3-4B-Base to match official Qwen3-4B-Instruct on held-out tasks.
KV4-RL uses selective INT4 KV quantization to speed Qwen3 GRPO rollouts nearly 2x while keeping benchmark accuracy.
SCAPO uses semifactual token stability to beat GRPO on AIME for two Qwen3 base models.
SCAPO adds semifactual token stability to GRPO, lifting Qwen3 AIME accuracy by up to 5.63 points.
SCALE gates frozen SFT features with entropy-trained controls, improving math and code scores on three Qwen models.
Self-play search distillation raises Qwen3-4B-Base's math benchmark mean from 24.1 to 36.6.
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.
Today 121 readers · 129 views · 26 reads · all time 199K views