Episode
HyperEyes: Dual-Grained Efficiency-Aware Reinforcement Learning for Parallel Multimodal Search Agents
- Podcast
- Daily Paper Cast
- Published
- May 12, 2026
- Duration seconds
- 1521
- Processing state
not_requested- Canonical source
- https://share.transistor.fm/s/3d74d9c7
Actions
POST https://stenobird.com/v1/public/podcasts/daily-paper-cast-7079649/episodes/hypereyes-dual-grained-efficiency-aware-reinforcement-learning-for-parallel-multimodal-search-agents/transcription-requests
Idempotently request low-priority transcript generation for this episode.GET https://stenobird.com/podcast/daily-paper-cast-7079649/hypereyes-dual-grained-efficiency-aware-reinforcement-learning-for-parallel-multimodal-search-agents.md
Read the agent-friendly Markdown representation of this episode resource.
Summary
🤗 Upvotes: 57 | cs.LG, cs.AI Authors: Guankai Li, Jiabin Chen, Yi Xu, Xichen Zhang, Yuan Lu Title: HyperEyes: Dual-Grained Efficiency-Aware Reinforcement Learning for Parallel Multimodal Search Agents Arxiv: http://arxiv.org/abs/2605.07177v1 Abstract: Existing multimodal search agents process target entities sequentially, issuing one tool call per entity and accumulating redundant interaction rounds whenever a query decomposes into independent sub-retrievals. We argue that effective multimodal agents should search wider rather than longer: dispatching multiple grounded queries concurrently within a round. To this end, we present HyperEyes, a parallel multimodal search agent that fuses visual grounding and retrieval into a single atomic action, enabling concurrent search across multiple entities while treating inference efficiency as a first-class training objective. HyperEyes is trained in two stages. For cold-start supervision, we develop a Parallel-Amenable Data Synthesis Pipeline covering visual multi-entity and textual multi-constraint queries, curating efficiency-oriented trajectories via Progressive Rejection Sampling. Building on this, our central contribution, a Dual-Grained Efficiency-Aware Reinforcement Learning framework, operates at two levels. At the macro level, we propose TRACE (Tool-use Reference-Adaptive Cost Efficiency), a trajectory-level reward whose reference is monotonically tightened during training to suppress superfluous tool calls without restricting genuine multi-hop search. At the micro level, we adapt On-Policy Distillation to inject dense token-level corrective signals from an external teacher on failed rollouts, mitigating the credit-assignment deficiency of sparse outcome rewards. Since existing benchmarks evaluate accuracy as the sole m…