agentic-training

Tag

Cards List
#agentic-training

@QwenAIFans: Four Claude Agents Train a Raw Qwen3-4B to Official Instruct Level — Athanor Paper from ByteDance and Others: 4 Claude Opus 5.5 Agents, Each on an 8-GPU Machine, Do Post-Training on Their Own for 104 Hours — No Human Teachers Allowed

X AI KOLs Timeline ↗ · 2026-10-21 Cached

The Athanor paper from ByteDance and other institutions shows 4 Claude Opus 5.5 agents, each occupying an 8-GPU machine, autonomously completing post-training in 104 hours with no human intervention, raising Qwen3-4B-Base to official Instruct level (dev set 15.9→67.3, held-out 9.2→48.2).

0 favorites 0 likes
#agentic-training

WorkRover: Scaling Long-Horizon Cowork Agent Training Via Recursive Environment Evolution

Hugging Face Daily Papers ↗ · 2026-10-13 Cached

WorkRover is a framework for training long-horizon cowork agents through recursive environment evolution, introducing Factorial Compatibility Replay and Verifier-to-Turn Credit Assignment to improve training signals. Post-training Qwen3.5-35B-A3B into WorkRover-35B-A3B yields large gains on JobBench, OfficeQAPro, and GDPvalv2 benchmarks.

0 favorites 0 likes
#agentic-training

@maximelabonne: Such a cool collab, very happy to see this kind of recipe getting open-sourced Train LFM2.5-2.6B on all the harnesses!

X AI KOLs Timeline ↗ · 2026-10-01 Cached

Maximilien Labonne highlights an open-sourced guide to training models with RL inside real agent harnesses like Claude Code, Codex, and OpenCode, enabling training of any model (e.g., LFM2.5-2.6B) on any task set across harnesses.

0 favorites 0 likes
#agentic-training

DeepSeek Elastic Compute (DSec): Sandbox Infrastructure for Effective Agentic Training at Scale

Lobsters Hottest ↗ · 2026-09-23 Cached

DeepSeek Elastic Compute (DSec) is a sandbox infrastructure for effective agentic training of large language models at scale, featuring elastic execution with unified SDK and integration with reinforcement learning frameworks.

0 favorites 0 likes
#agentic-training

@Azaliamirh: Check out TRACE, a new self-improvement approach where the agent identifies the missing capabilities behind its own fai…

X AI KOLs Timeline ↗ · 2026-07-09 Cached

TRACE is a new self-improvement approach where an AI agent identifies the missing capabilities behind its own failures and trains itself to address them. TRACE-trained Qwen3.6-27B achieves 73.2% on SWE-bench Verified, outperforming much larger models with fewer training rollouts.

0 favorites 0 likes
← Back to home

Submit Feedback