Mis à jour quotidiennement1794 histoires vérifiées
Couverture par l'IA vérifiée à la source des lancements de produits, des changements de politique, de la recherche sur la sécurité et des évolutions de l'industrie, expliquée dans un anglais simple par une équipe éducative à but non lucratif.
Chaque article est lié aux preuves disponibles les plus solides : sources originales lorsqu'elles sont disponibles, sinon reportages clairement attribués.
Que s’est-il passé, pourquoi c’est important et que regarder – sans le jargon.
Lorsque le signal est faible, nous ne publions rien plutôt que de compléter le flux.
Histoires d'IA vérifiées à la source, les plus récentes en premier, pour les personnes qui ont besoin de comprendre l'IA sans courir après le battage médiatique.
A five-author ICML 2026 paper proposes measuring changes in nonlinear input interactions, not just final answers, when small prompt edits destabilize language models.
A new arXiv preprint studies whether reasoning models can learn tasks sequentially without losing ground, and proposes Continual Prompt Replay to close the gap with joint multitask training.
A new arXiv preprint argues that sequence-pooled normalization can provide global context even when a convolutional sequence labeler has a short receptive field. The finding could affect how researchers assess streaming limits, architecture design and attribution results.
A single-author preprint reports that Transformer and Mamba architectures develop similar near-marginal slow-mode dynamics despite using different internal mechanisms. The finding remains unverified beyond the supplied arXiv record.
A new arXiv preprint reports that strong inference-time guidance can push protein language models toward degenerate sequences that still score well under the optimized property oracle. The authors propose a low-cost post-hoc filter based on the model’s natural activation statistics.
A paper accepted by IEEE ICDM 2026 reports that adaptive inversion reconstructs substantially more of the original text from Gaussian-noise-protected embeddings than an existing generative baseline, challenging a widely used privacy assumption without establishing real-world exposure or independent replication.
A single-author arXiv preprint introduces the Middle East Cultural Sensitivity Score and reports that two language models reproduce structural Orientalist patterns in conversations about the region. The findings are preliminary and have not been independently verified.
A one-author arXiv preprint claims that a direction derived from nine emotion categories and narrative examples can track positive-negative valence in text, images, audio and brain recordings, while showing clear limits across concepts and model families.
A new arXiv preprint describes DeepTCM1.0, a system using 11 specialized AI agents and iterative review to analyze possible mechanisms of traditional Chinese medicine formulas. It reports an evaluation design but no numerical results, clinical validation, or evidence of improved patient care.
An arXiv-listed Ph.D. dissertation studies how backdoor attacks can affect language models and vision-language models, including methods for analysis, detection and attack design.
An arXiv preprint reports that four large language models rated candidate profiles more favorably when linked to higher-prestige institutions and journals. The authors say institutional and geographic cues can shape evaluation, while the tested models and professional domains remain unspecified in the supplied source.
An arXiv paper reports that three of four tested language models shifted resource-allocation probabilities differently when the same clinical scenario was evaluated with or without the model’s prior response in context.
Un briefing utile chaque semaine
Recevez les actualités vérifiées de la semaine sur l'IA, les données originales, les outils utiles, les choix d'apprentissage et les nouveaux travaux d'IA.
Embaucher un professionnel de l’IA ou lancer un produit d’IA utile ? Présentez-le aux gens qui sont venus ici pour apprendre et agir.
Publier une offre d'emploi IA Soumettre un outil d'IA