AI
Shared Circuits for Shared Grammar: Tracing Subject-Verb Agreement Across Languages
Researchers have found that multilingual large language models (LLMs) share similar internal mechani...
AI
Beyond LLM-Based Reasoning: Lightweight GNNs for Agent Failure Attribution
Researchers have proposed a new method for identifying the causes of failures in multi-agent systems...
AI
Compress and Forget: bitsandbytes Quantization Amplifies Proactive Interference in LLMs
Researchers have found that using a specific type of quantization called bitsandbytes can amplify pr...
AI
TranslatePsy-AfriSLM: High-Quality Data Scaling For Low-Resource Machine Translation
Researchers have created a collection of open-source machine translation resources for 19 Sub-Sahara...
AI
X2Streaming-TTS: Causal Token-Level Text-to-Speech from Streaming Text with Speech-State Inheritance
Researchers have developed a new text-to-speech system called X2Streaming-TTS th...
AI
Learning What to Fail On: Failure-Mode Contextual Bandits for Adversarial Data Curation
Researchers have developed a new method for improving the robustness of natural ...
AI
Execution-grounded evaluation reveals hidden failures in language-model calculations for environmental science
Researchers have developed a new benchmark called AtmosCoder-Bench to evaluate t...
AI
Learning Canonical Register Automata over Ordered Data Domains
Researchers have developed a new method for training deterministic register auto...