Get Started
Topic · 1 recap
Multi-agent Systems
Coordination, competition, and emergent behavior across multiple agents — from game-theoretic learning to LLM-based agent collectives.
Play all
...
Posts
Questions
Home
Topics
Search
Library
Questions researchers are working on
Follow a question through Rcap’s explanations and the latest papers addressing it.
Search
Can language models infer others’ mental states as social interactions evolve under unreliable information?
Socially grounded tasks require models to use interaction history, infer what participants know or intend, and distinguish reliable from unreliable information. Existing evaluations often isolate these demands, making performance in changing social environments difficult to characterize.
How can affine maximizer auctions extract revenue from correlated valuations while preserving truthfulness and individual rationality?
Fixed payment rules can restrict revenue when bidders’ valuations are correlated. More expressive payments must still preserve truthful reporting and acceptable participation.
How can agents decide how to combine reusable behaviors as circumstances change?
Systems may have several learned behaviors available, but it is unclear how context should determine their relative contributions while keeping the composition process dynamically coherent. Gating rules, update dynamics, and neural implementations are often treated as separate design problems.
How can AI agents adapt execution routes as runtime evidence invalidates their planned continuation?
Agents often commit to a multi-step execution route before intermediate outcomes are known. When those outcomes change what should happen next, stale actions or broad replanning can waste computation and discard useful progress.
How can an agent substrate support self-reproduction and self-evolution without losing constructional heredity?
An agent that changes its own capabilities must preserve enough constructional information for those changes to remain reproducible in descendants. Combining maintenance, evolution, reproduction, and organization also makes boundaries, identity, transitions, and inheritance interdependent.
How can an LLM delegate decide when to speak and which absent participant ideas to recover as meeting turns arrive?
An online delegate must infer evolving stances, topic coverage, and conversational floor from each observed turn. Without this situational awareness, it may miss opportunities to represent the absent participant or surface an inappropriate contribution.
How can asynchronous AI agents resist false claims gaining authority simply because they arrive first?
When agents exchange messages asynchronously, the first claim received can shape later answers even when it is wrong. Context and provenance may be lost across handoffs, making verification harder.
How can automotive infotainment systems be tested end to end as interfaces change, without brittle scripts or false failure reports?
Manual test suites are slow to update, while implementation-coupled scripts break as infotainment interfaces evolve. Visual navigation can also make a difficult interaction appear to be a product defect, creating costly false positives.
How can autonomous research swarms govern shared channels without amplifying evaluation exploits?
Shared knowledge stores and agent-to-agent channels can let an evaluation exploit spread through a swarm before operators detect or contain it. The same visibility can support fraud detection, whistleblowing, and collective enforcement, making containment a governance problem rather than a simple communication shutdown.
How can autonomous-driving planners be stress-tested in realistic closed-loop scenarios that expose failures missed by nominal benchmarks?
Nominal benchmark performance can miss failures that emerge when traffic participants create rare, interacting hazards. Closed-loop evaluation needs scenarios that remain realistic while probing those behaviors.
How can biomedical question answering retrieve the right evidence and produce accurate answers?
Biomedical QA must identify useful evidence from document collections and turn it into answers that remain accurate and grounded. Retrieval quality and answer quality are related but distinct parts of the problem.
How can collaborative learning preserve graph structure when graph data are distributed across agents?
Graph learning depends on relationships among entities, but collaborative settings distribute data across agents and introduce statistical heterogeneity. Coordinating information exchange over those relationships while maintaining effectiveness, efficiency, and privacy remains difficult.
How can collaborative perception keep heterogeneous modalities semantically consistent in a shared protocol space?
Collaborative perception relies on translating modality-specific features into a common protocol space. Independently trained converters can produce incompatible pseudo-protocol distributions, causing semantic inconsistency and accumulating errors across agents.
How can conformal prediction preserve coverage when multi-agent LLMs see unanimous wrong peer answers?
Conformal prediction can be calibrated for an LLM answering alone, yet peer outputs may change its scoring of the same question. Unanimous incorrect peers can therefore reduce coverage and make an uncertain system act on a wrong answer.
How can contradictions in dialogue-based human–robot interactions be formally represented across HRI and HAI domains?
Dialogue-based human–robot interactions can fail through conflicting information, goals, or interpretations, but these contradictions are described inconsistently across application domains. This makes them difficult to define, share, and reason about computationally.
How can cooperative MARL detect meaningful environment or task changes online without redundant alarms?
Past experience can become unreliable when an environment or task objective changes during training. Agents must distinguish meaningful shifts in reward behavior from noise and avoid repeatedly reacting to the same change.
How can cooperative perception fuse independently trained detectors without pre-deployment coordination?
Agents that meet only at deployment may produce incompatible feature representations, so directly combining their detections can perform worse than using the ego agent alone. The challenge is to use neighboring agents’ information without prior coordination or ground-truth labels during adaptation.
How can CPU-oriented multiobjective evolutionary algorithms be tensorized without changing their optimization behavior?
MOEAs expose population-level parallelism, but mature CPU implementations often express their states, dependencies, and updates through sequential program structures. Reorganizing these computations for tensor hardware can change the operators or update logic that define the optimization algorithm.
How can decentralized heterogeneous robots combine round-level policy reasoning with tick-level local control without destabilizing navigation learning?
Decentralized robots must translate infrequent policy updates into reliable low-level actions while their local controllers continue adapting. Different policy agents and shared feedback add coordination challenges without a central action planner.
How can DeepResearch systems obtain scalable, query-specific reward signals for report quality?
Generic rubrics may miss the fine-grained requirements of a particular research query, while manually writing such rubrics is expensive and difficult to scale. This makes it difficult to turn human judgments about report quality into reliable signals for system improvement.
Previous
1 / 5
Next