build-with-paper

Here are 6 public repositories matching this topic...

loversky02 / super-agent

Unified cost-aware router: model x reasoning-depth on one Thompson bandit (factored policy), bi-temporal forget-on-drift statistics memory + auto-drift proxy, GRPO-internalized Qwen3-4B depth policy.

reinforcement-learning thompson-sampling multi-armed-bandit llm-agents llm-routing agent-memory grpo cost-aware-routing build-with-paper

Updated Jul 9, 2026
Python

loversky02 / agent-memory-lab

Star

Runnable lab measuring invalidation/staleness in agent memory — paper: Are We Ready For An Agent-Native Memory System? (arXiv:2606.24775)

benchmark invalidation mlx memory-systems bi-temporal rag paper-reproduction llm kuzu llm-agents agent-memory agentic-evaluation build-with-paper

Updated Jun 26, 2026
Python

loversky02 / System-III-Router

Star

Learned deliberation routing (System III Configurator) from 'Critique of Agent Model' (arXiv:2606.23991): a cost-aware bandit over reasoning depths (direct/cot/plan+verify), live on Vi-GSM8K, with a GRPO path for Qwen3-4B.

multi-armed-bandit reasoning vietnamese-nlp deliberation llm-agents llm-routing grpo cost-aware-routing build-with-paper

Updated Jul 2, 2026
Python

loversky02 / vi-gsm8k-agentic

Star

Vietnamese GSM8K agentic self-instruct dataset — proven to beat machine-translated data (Qwen3-4B: 81.0% vs 76.5%)

dataset vietnamese-nlp synthetic-data instruction-tuning gsm8k self-instruct math-reasoning build-with-paper

Updated Jun 27, 2026
Python

loversky02 / spiced-mini

Star

Hands-on, runnable demos of SPICED (self-play + a pinch of human data, arxiv 2606.19370) at tiny scale on a Mac M5: demo-regularized self-play, the RLHF KL knob, and LoRA on a real LLM.