Tencent ContextPilot Agent
ContextPilot offers a more efficient approach to long-context tasks compared to existing methods, potentially reducing computational costs for developers in the region.
Tencent developed ContextPilot, an agent that manages its own working context with global planning, long-term memory and adaptive soft compression. The model uses context and entropy variation to identify which editing decisions matter in long-horizon tasks. ContextPilot beats existing baselines across several base models on long-context QA and deep search while maintaining a more compact working context. The training code is available on arXiv.
Interesting paper from Tencent. Tencent trains an agent to manage its own working context, and assigns credit at the level of individual context edits. Long-horizon tasks force a model to retrieve, integrate and maintain scattered information across many turns, and keeping every interaction history makes the working context grow without bound. Recent proactive methods let a model edit its own context with tools, but the toolset stops at search, deletion and summarization. ContextPilot adds global planning, long-term memory and adaptive soft compression, so the agent can offload information rather than only discard it. The training side is where it gets interesting. Standard RL hands the final trajectory reward to every intermediate edit equally. ContextPilot uses context and entropy variation to find which editing decisions actually mattered, samples branches at those points, and estimates action-level advantages from all branched trajectories passing through that edit. On long-context QA and deep search it beats existing baselines across several base models while holding a more compact working context. Code is available. Paper: arxiv.org/abs/2608.28476 Chat with Paper: academy.dair.ai/papers/context… 💬 7 🔄 2 ❤️ 17 👀 2238 📊 10 ⚡
