Researchers propose TTSR, a test-time training framework where a single LLM alternates between Student and Teacher roles to improve its own reasoning. The method generates targeted variant questions based on analyzed failure patterns, showing consistent improvements across mathematical reasoning benchmarks without relying on unreliable pseudo-labels.
Researchers have introduced OSCAR, a query-dependent compression method for Retrieval-Augmented Generation that speeds up inference 2-5x while preserving accuracy. Unlike traditional approaches, OSCAR compresses retrieved information dynamically at inference time rather than offline, eliminating storage overhead and enabling higher compression rates.
Researchers propose LaDiR, a framework that replaces traditional autoregressive decoding with latent diffusion models to improve LLM reasoning. The approach encodes reasoning steps into compressed latent representations and uses bidirectional attention to refine solutions iteratively, enabling parallel exploration of diverse reasoning paths.
Researchers have introduced VeriStruct, a framework that extends AI-assisted formal verification from individual functions to complete data structure modules in Verus. The system successfully verified 128 of 129 functions (99.2%) across eleven Rust data structure modules by using a planner module to generate abstractions, type invariants, and proof code, with automatic error correction for Verus syntax.
Researchers from ETH Zurich and Anthropic have demonstrated that pseudonymous internet users can be de-anonymized using commercially available AI models at a cost of just a few dollars per person. The attack works in minutes and calls fundamental assumptions about online anonymity into question.
Google announced Gemini 3.1 Pro, positioning the model for complex problem-solving tasks requiring deeper reasoning than previous versions. The release follows Gemini 3 Pro (November 2025) and Gemini 3 Flash (December 2025).