Adversarial Concept Search: Predicting Compositional Errors From Feature Geometry
Quick Answer
This paper introduces a method to predict compositional errors in LLMs by analyzing their representational geometry.
Quick Take
This paper introduces a method to predict compositional errors in LLMs by analyzing their representational geometry. By identifying interference between salient features, the approach anticipates failure modes in tasks like multihop reasoning and multilingual recall, enabling targeted stress tests and enhancing active learning for real-world applications.
Key Points
- Predicts LLM failure modes using representational geometry without specific input evaluation.
- Identifies interference between features as a cause of compositional failure.
- Demonstrates reliable composition in near-orthogonal concept pairs.
- Applicable to tasks requiring systematic composition like multilingual factual recall.
- Lays groundwork for scalable active learning and stress testing in deployment.
Paper Resources
Article Excerpt
From source RSS / original summaryarXiv:2606. 13934v1 Announce Type: new Abstract: Humans cannot always intuit what scenarios are most challenging to LLMs. Hoping to capture challenging edge cases, developers either design problems to be difficult for humans or curate extensive benchmarks. What if we could instead anticipate which scenarios a model will fail on? In this paper, we use an LLM's representational geometry to predict which concept combinations it will fail on.
We attribute this compositional failure to interference between salient features. In tasks that require systematic composition - toy programmatic settings, multihop reasoning, multilingual factual recall - we find that when a pair of concepts is encoded near-orthogonally, the model reliably composes them. When their linear encodings are close, producing interference, the model fails to compose them.
Our method reliably anticipates failure modes across different compositional tasks, without evaluating specific inputs. These results lay the groundwork to use representational geometry to identify high-risk examples, construct targeted stress tests, and provide a scalable foundation for active learning in real-world deployment.
Reader Mode unavailable (could not extract clean content).
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →Arbor: Tree Search as a Cognition Layer for Autonomous Agents
Arbor introduces a multi-agent framework utilizing structured tree search for optimizing LLM inference, achieving up to 193% throughput-latency improvement compared to vendor-optimized systems. It employs an Orchestrator and Critic agent for stability and coordination, demonstrating hardware-agnostic performance with minimal variance.