Model releases, architecture work and prompting research on large language models — from frontier-lab announcements to the arXiv papers behind them.
arXiv:2609.22700v1 Announce Type: new Abstract: Step-level reasoning evaluators are commonly based on autoregressive language models, whose causal attention restricts each step representation to the...
arXiv:2609.22771v1 Announce Type: cross Abstract: Recent advances in Audio LLMs have achieved human-level speech recognition, yet existing systems struggle to capture paralinguistic aspects such as s...
arXiv:2609.22793v1 Announce Type: new Abstract: LLM-based machine translation evaluation can closely match human judgments, but in practice it remains largely diagnostic, with the signals rarely tran...
The paper investigates whether contrastive activation directions in large language models truly capture a functional valence axis or merely reflect correlated features of the contrast used to extract them. By conducting controlled interventions in a maze task across multiple LLM checkpoints, the authors show that directions fitted on one explicit outcome encoding transfer well to another, suggesting the readout is not tied to surface form. However, when the same outcome is achieved through announced versus unannounced histories, transfer degrades, indicating the direction remains strongly conditioned on prior announcements and does not represent a history‑invariant scalar valence state.
arXiv:2609.22884v1 Announce Type: new Abstract: Long-context inference has become a defining capability of large language models, but exact dense attention remains costly due to its quadratic scaling...
arXiv:2609.22934v1 Announce Type: new Abstract: Large language models (LLMs) increasingly mediate human decisions and communication, yet their behavioural regularities remain difficult to characteriz...
MolSC is a new dataset of 181,000 substituent-level examples that captures how attaching specific substituents to molecular scaffolds changes properties such as bioactivity and physicochemical descriptors. The authors also provide MolSC-Bench, a held‑out benchmark of 1,541 examples that are disjoint from MolSC at scaffold, substituent, and molecule levels. Experiments show that training molecular large language models on MolSC markedly improves their ability to predict substituent contributions, outperforming existing models on a range of downstream chemistry tasks.
arXiv:2609.23139v1 Announce Type: new Abstract: Most medical vision-language models (VLMs) excel at open-ended report generation and VQA but provide limited support for structured, fine-grained clini...
ValueDiff introduces a value‑geometric KV cache eviction strategy for large language models that suppress attention sinks. It ranks tokens by the L2 deviation of their value vectors from the cache mean, a score that aligns with minimal‑disturbance eviction under a max‑entropy assumption. Across several benchmarks—RULER, LongBench, and MATH‑500—ValueDiff consistently retains a higher proportion of useful tokens than prior methods, especially under tight cache budgets.
arXiv:2609.23321v1 Announce Type: cross Abstract: Low-rank adaptation (LoRA) has become a key technology for serving large-scale personalized large language models and diffusion models in the cloud....
arXiv:2609.23397v1 Announce Type: new Abstract: Shrimp diseases continue to cause devastating losses in the aquaculture industry, driving a critical need for robust, automated detection. This work co...
RPMem introduces a two‑stage architecture that compiles each session into a model‑independent latent memory and then consolidates it with retained memory via a task‑trained recurrent gate. The consolidated memory is mapped to backbone‑specific low‑rank adaptation (LoRA) parameters, enabling the memory to transfer when the backbone is replaced. Across three long‑term memory benchmarks and five diverse backbones, RPMem achieves broad generalization with near‑constant update cost and memory footprint, outperforming existing parametric and text‑based baselines on the PERMA benchmark.
arXiv:2609.23600v1 Announce Type: new Abstract: Prompt learning efficiently adapts vision-language models (VLMs) to downstream tasks, but gains on seen classes often come at the expense of generaliza...
arXiv:2609.23601v1 Announce Type: new Abstract: Long-video understanding must capture transient visual evidence under strict token budgets, yet existing methods compress frames, append memory tokens,...
arXiv:2609.23716v1 Announce Type: cross Abstract: Textual-gradient methods automate prompt optimization through natural-language feedback, but their iterative updates can be unstable. We identify two...
FLARE introduces a dense supervision paradigm for long‑horizon coding agents, leveraging a Generative Reward Model (GRM) trained via the RADAR diagnostic framework. The GRM provides real‑time, step‑level risk feedback, enabling FLARE to act as an active scaffold that intercepts high‑risk steps during inference and supplies structured signals for post‑training fine‑tuning and reinforcement learning. Experiments show FLARE outperforms existing methods, achieving a 5× reduction in token consumption and significant performance gains in both supervised fine‑tuning and RL settings.
arXiv:2609.23825v1 Announce Type: new Abstract: We present a comprehensive benchmark of Federated Learning (FL) for multilingual Automatic Speech Recognition (ASR), evaluating four Speech-LLM archite...
arXiv:2609.23853v1 Announce Type: new Abstract: Disaster-risk-reduction archives describe hazard events in prose that databases such as EM-DAT (Delforge et al., 2025) cannot ingest directly. We prese...
arXiv:2609.23980v1 Announce Type: cross Abstract: AI agents now report vulnerabilities faster than maintainers can review them. Reports often depend on security properties specific to the application...
arXiv:2609.24156v1 Announce Type: cross Abstract: Most existing time series forecasting methods rely solely on numerical observations, overlooking rich contextual information from auxiliary texts. Re...