Take three frontier mixture-of-experts models (Alibaba, OpenAI, NVIDIA; 3.6-4.0B active parameters each) and fine-tune them to reason in a low-resource language. On accuracy benchmarks almost nothing happens, and the benchmark itself is noise at this scale: changing only the random seed moves the score by 7.7 points, more than every data and recipe effect we measured. That null is our first result. The real changes live where accuracy cannot see. Base models never think in Greek: 0 of 1,000 reasoning traces, even when the question is Greek, so the model answers correctly while reasoning in a form its user cannot read, audit, or correct. After supervised fine-tuning (SFT), every released checkpoint reasons in the language of the question on ~98% of items, one family at 3x fewer tokens, with judged grammaticality improving on all four models and general ability within a few points of each base: nothing was forgotten, and fluency was gained. We propose six behavioural dimensions that make such changes measurable, each gated to reject any metric that correlates with output length, and we report how our own instruments lied: six failures, each caught by a control. What SFT cannot do is fix its own defects: a quarter of answers skip the requested format, answers leak into the reasoning channel, and an explicit"think in English"is obeyed under half the time. Reinforcement learning with verifiable rewards, pre-registered before training, fixes the first two outright (fallback 24% to 2.5%, leak 3.5% to 0.0%, both against a flat random-reward control) and moves the third (+9.1pp), while the Greek reasoning habit survives an accuracy-only gradient untouched. We release five checkpoints. The instruments, the controls and the pre-registration travel to any low-resource language; Greek is the case that let us measure them.
This work proposes MemCatalyst, a set of data poisoning tools, aiming to amplify the data auditing performance on VLMs, and forces VLMs to over-learn specific inconsistencies between image features and textual semantics during training, thereby increasing their susceptibility to membership information auditing.
Xukun Luan, Jinyan Liu, Yuhui Gong et al.· 0 citations
InnerExpert is introduced, the first method to leverage MoE-specific signals for per-token hallucination detection, and shows that InnerExpert outperforms existing methods across five datasets and two MoE architectures.
João Fonseca, Rodrigo Rodrigues, Paolo Romano· 0 citations
A reinforcement learning-based framework for robotic grasp refinement, integrating keypoint-based object representations with a Deep Q-Network (DQN), is proposed, offering a scalable and adaptable solution for contact-rich manipulation tasks.
Amir Arsalan Nematollahi, Shayan Ahmadi, M. T. Masouleh et al.· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
MoNe is a lightweight modular neural memory that attaches to any frozen pretrained Transformer to enable long-context inference without retraining, achieving strong performance on needle-in-a-haystack and word extraction benchmarks from RULER, where ICL degrades sharply.
Won-Yong Cho, Kyubyung Chae, Tribhuvanesh Orekondy et al.· 0 citations
This work employs encoder-decoder networks built around long-short term memory (LSTM) cells in a distributed optimization algorithm that allows unprecedented reconstruction accuracy or the usage of different prediction-horizon lengths without the necessity to retrain.
Torben Schiz, P. H. Nardelli, Henrik Ebel· 0 citations
This analysis identifies a common obstruction: cheap nuisance interpolation causes the refit to underweight the truly predictive coordinate, and an exact target-mass identity and a two-sign argument turn this effect into clipped prediction loss.
A novel mechanism to automatically identify which of these point-labels are suitable, and which are actively harmful, when used for propagation is introduced, paving the way for scalable ecological analysis.
César Borja, Breck A. McCollum, Jarrett E. K. Byrnes et al.· 0 citations
Reflex-Guard is introduced, a lightweight guardrail that runs locally that uses jailbreak-aware preprocessing, compact sentence-transformer embeddings, and seven fast binary classifiers that enable high-accuracy prompt safety filtering with much lower latency than existing solutions.
Istiaque Ahmed, Afia Anjum Borsha, Ranat Das Prangon et al.· 0 citations
Spaced Repetition Training (SRT) is introduced, a continual learning framework inspired by cognitive science, which schedules sample-rehearsal using the SuperMemo-2 (SM-2) algorithm, and preserves broad benchmark performance that naive continual pre-training and uniform replay substantially degrade.
Results indicate the visual component is dominant but not solely responsible for skill prediction; further work is needed to conclusively disentangle transferable skill features from those bound to a specific visual domain.
Hanna Hoffmann, F. Bechtolsheim, Stefanie Speidel et al.· 0 citations
The Structural Gap Hypothesis Agent (SGHA), a fully automated, corpus-first research-problem discovery system that runs entirely on a local LLM, is introduced and compared with the AI Scientist-v2 idea formulation module in five machine-learning domains.
A new machine-learning framework aims to improve the success rate of computational protein design while moving away from results that reproduce sequences found in nature.
MIT News · Artificial Intelligence· news.mit.eduAug 24, 2026
A new method for surgically removing training examples from a model reveals that as datasets grow, the link between what a model learns and what it produces dissolves.
