Aleph Alpha released Kolibri on October 3, 2026, a 78.1-billion-parameter open-weight Mixture-of-Experts model with a context window of up to 1,048,576 tokens, trained on 768 NVIDIA B200 GPUs in Germa…
A self-supervised method called Reflective Recovery, which turns failed reasoning trajectories into recovery training data, raised accuracy on DeepSeek-R1-Distill-Qwen-7B from 30.0% to 37.5% on AIME 2…
Researchers proposed OBC-Prune, an outcome-based calibration method for pruning large reasoning models that weights reasoning tokens by their causal contribution to correct answers rather than by unif…
A new arXiv study introduces the Token Economy Score (TES), a metric measuring the accuracy gain of reasoning-capable large language models over non-reasoning baselines normalized by token multiplier,…
Microsoft introduced MAI-Thinking-1, a 35B-active, ~1T-total parameter sparse Mixture of Experts reasoning model that matches leading models on software engineering benchmarks and is preferred to Sonn…
A new technique called KV-Cache Grafting boosts a frozen Gemma-4-12B model from 80.0% to 93.3% on AIME 2025 without changing any weights, according to a paper submitted to arXiv on 15 Jul 2026. The me…
Researchers introduced TREK, a method that improves AI problem-solving by using verified output trajectories to extend model learning. TREK boosted Qwen3-8B's performance on AIME 2024 from 36.9 to 40.…
Researchers introduced VeryTrace, a zero-shot verification-and-repair framework that formalizes Chain-of-Thought reasoning traces into a compilable Domain-Specific Language, enabling step-level error …
OpenAI released an update to GPT-5.5 Instant on May 28, improving response style, readability, and personalization, with the model scoring 81.2 on the AIME 2025 math test versus 65.4 for the prior ver…
Researchers introduced Sequential Bayesian Belief Tracking (SBBT), a framework that estimates the likelihood of a correct final answer from partial reasoning traces by calibrating observation likeliho…