arXiv — cs.AI preprintsInternational16 October 2026
Amplitude-Only FFN Intervention for Tool-Structured LLM Inference: Method, Gated Evaluation Protocol, and Cross-Model Empirical Results
This is an official announcement record
Firsthand records what arXiv — cs.AI preprints announced and links to the original. The wording below is theirs, not ours.
arXiv:2607.11183v4 Announce Type: replace-cross Abstract: We study Amplitude Gating (AG), an inference-time intervention that masks feed-forward network (FFN) activation coordinates while keeping pretrained weights fixed. Six SwiGLU sites separate shared and gate-only inputs, expanded gate/value channels, and the FFN output. Magnitude-ranked entropy masks and a Newton-Schulz windowed ranking variant provide alternative forward paths without retraining. The corrected September collection covers three Qwen checkpoints and eight text/tool endpoints. Under frozen local text scorers, the displayed
Read the official announcement
Opens arxiv.org
More from arXiv — cs.AI preprints
- The World Won't Wait: Evaluating and Teaching Agents Under a Live Clock16 October 2026
- Evaluating Model Capabilities After Benchmark Saturation16 October 2026
- Value-Informed Schedule Trimming for Accelerated Sequential Monte Carlo Guidance in Discrete Diffusion16 October 2026
- AlphaDiana: A Measurement System for Agents in Verifiable Reasoning Tasks16 October 2026
- When Correctness Feedback Erases Evidence: Phase-Separated Completeness Auditing in Verus16 October 2026
This content is for informational purposes only and is not professional advice. Specifications, prices, plan tiers, and features change frequently and may differ from what is shown here; verify current details on the manufacturer's or company's official page before purchasing. Ratings are based on analysis of published documentation, not independent lab testing.