Research map / Architectures and efficiency / RNNs and normalization
Chanakya: Learning Runtime Decisions for Adaptive Real-Time Perception
Anurag Ghosh, Vaibhav Balloli, Akshay Nambi, Aditya Singh, Tanuja Ganu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| learned Chanakya policy (R2) instead of best static policy Streamer s=600 | Faster R-CNN + Kalman Filter · Argoverse-HD | sAP: 20.4 -> 21.3 | Improved |
| learned policy instead of AdaScale dynamic traditional policy | Streamer + Kalman Filter · Argoverse-HD | sAP: 13.8 -> 21.3 | Improved |
| add modified shrinking tail scheduler to Streamer + AdaScale (ablation) | Streamer + AdaScale · Argoverse-HD | sAP: 13.4 -> 13.8 | Improved |
| expert tuning of proposals (np=500) on Streamer s=600 | Streamer · Argoverse-HD | sAP: 20.4 -> 20.9 | Improved |
| use R2 (advantage over fixed policy) instead of R1 reward (ablation) | Chanakya · Argoverse-HD | sAP: 21.0 -> 21.3 | Improved |
| use R2 instead of R1 reward (ablation) | Chanakya · ImageNet-VID | sAP: 37.5 -> 37.2 | Hurt |
| use R2 instead of R1 reward on Xavier NX (ablation) | Chanakya · ImageNet-VID Xavier NX | sAP: 19.4 -> 22.4 | Improved |
| learned Chanakya policy instead of static policy on edge device | Chanakya · Argoverse-HD Xavier NX | sAP: 7.8 -> 8.3 | Improved |
| learned policy instead of static FCOS policy | Chanakya with FCOS · Argoverse-HD | sAP: 16.7 -> 20.7 | Improved |
| learned policy on V100 GPU with expanded decision space | Chanakya · Argoverse-HD V100 | sAP: 18.3 -> 27.0 | Improved |
| add tracker and enlarge decision space to 500 configurations | Chanakya · Argoverse-HD | sAP: +1.6 sAP | Improved |
| drop one context feature (ablation) | Chanakya · Argoverse-HD | sAP: approx 1 sAP | Hurt |
Closest papers by meaning
- Unleashing Perception-Time Scaling to Multimodal Reasoning Models (ICLR 2026 · similarity 0.84)
- FalconPerception-HD: High Density Perception via Reinforcement Learning (NeurIPS 2026 · similarity 0.83)
- DeepEyes: Incentivizing "Thinking with Images" via Reinforcement Learning (ICLR 2026 · similarity 0.83)
- ACTIVE-o3 : Empowering MLLMs with Active Perception via Pure Reinforcement Learning (ICML 2026 · similarity 0.83)
- SPARC: Separating Perception And Reasoning Circuits for Test-time Scaling of VLMs (ICML 2026 · similarity 0.83)
- The Perceptual Bandwidth Bottleneck in Vision-Language Models: Active Visual Reasoning via Sequential Experimental Design (ICML 2026 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
