Hardware
Banger paper from NVIDIA on long running agents. A model can accept 128K tokens of context and still make more mistakes the longer it works …
Banger paper from NVIDIA on long running agents. A model can accept 128K tokens of context and still make more mistakes the longer it works through a task. If your agent loses its place partway throug
Banger paper from NVIDIA on long running agents. A model can accept 128K tokens of context and still make more mistakes the longer it works through a task. If your agent loses its place partway through a long table or ledger, this work measures what causes it. The setup: Long-Transduction asks a model to keep reading, updating and outputting state-dependent results over thousands of outputs, and varies three factors separately. Results: Across seven open-weight models, accuracy drops 62.8% when context grows from 4K to 128K, 36.5% when only the input format changes, and 39.9% when the per-step operation gets harder. Paper: https://academy.dair.ai/papers/staying-on-task-testing-the-foundations-of-long-horizon-agent-reliability-2609.38712
Related
- Good implementation of agent shared memory from the NVIDIA team.
- Banger paper from NVIDIA. Agentic reasoning needs models that are not just capable, but efficient at long-context inference. The agent model…
- KVCMAS: Efficient KV cache Correction for Shared Context in Multi-Agent Systems
- RIS-Kernel: A Model-Agnostic Architecture for Long-Context LLM Inference via Sparse Attention
Source: DAIR.AI (X) | 2026-10-02