AutoTool: New Framework Enables LLM Agents to Dynamically Select and Adapt to Changing Tool Sets
Researchers have introduced OmniMem, a streaming framework designed to reduce memory bottlenecks in audio-visual large language models processing long-form video. Current models struggle because video and audio tokens grow linearly, overwhelming key-value caches during inference. OmniMem addresses this by separately managing visual and audio memory and using perturbation-aware selection to retain the most informative states, yielding measurable accuracy gains on standard benchmarks.
Audio-visual large language models (LLMs) show strong potential for understanding long videos, but their practical deployment is constrained by the linear growth of video and audio tokens and the associated key-value (KV) cache memory demands. OmniMem tackles this with a modality-aware memory allocation strategy that manages visual and audio contexts separately, addressing the token imbalance that arises because the two modalities generate very different volumes of data. A perturbation-aware memory selection mechanism identifies and retains the most informative, non-redundant KV states, enabling compact memory without degrading long-range comprehension. The framework also incorporates optional budget-aware fine-tuning, which trains the model to consolidate useful information into the retained memory slots. Evaluated on VideoMME Long, LVBench, and LVOmniBench using the video-SALMONN 2+ and Qwen-2.5-Omni model families, OmniMem consistently outperforms strong training-free compression baselines by 2–4 percentage points in absolute accuracy under equivalent memory budgets, with a further 1–2 point improvement after fine-tuning. Code has been made publicly available by the authors.
What's missing
The study does not report wall-clock inference latency or throughput measurements, leaving it unclear whether the perturbation-aware selection process introduces meaningful computational overhead at deployment time. Additionally, evaluations are limited to two model families; generalizability to other audio-visual LLM architectures remains untested.
What different sources said
- arXiv cs.CLCenter
TLRD: Teaching LLMs to Reason over Tabular Data with Tri-Level Rationale Distillation
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.