New mRNA immunotherapy eliminates pancreatic tumors in mice

Good news! Cancer is history (soon)!

"... have developed an immunotherapy using a cocktail of messenger RNAs (mRNA) that could potentially transform pancreatic cancer treatment.

The study ... combines immune cytokine and tumor-associated antigen mRNAs into a single injection to treat pancreatic ductal adenocarcinoma.

Approximately 50% of mice with pancreatic cancer treated with the mRNA immunotherapeutic cocktail had complete tumor responses and, more impressively, remained disease-free for a year, even after treatment had stopped. 

"It's unheard of to get a response like this in these models of pancreatic cancer," ..."

From the abstract:

"Immunotherapy has limited success in pancreatic ductal adenocarcinoma (PDAC) due to an immune exclusive tumor microenvironment (TME) that lacks many cytokines necessary for Natural Killer (NK) and T cell responses.

Here, we design multiplexed mRNAs encoding interleukins, chemokines, and interferons as a safe and effective cytokine therapy for PDAC.

Intratumoral injection of IL-12, IL-18, CCL5, CXCL10, and IFNβ mRNAs achieves robust yet transient cytokine expression, leading to NK and CD8+ T cell activation and reduced tumor growth and fibrosis in PDAC transplant mouse models.

Combining cytokine with tumor antigen mRNAs enhances dendritic cell antigen presentation and CD8+ T cell priming locally and systemically that prolongs animal survival after a single dose.

Remarkably, nanoparticle encapsulation of the cytokine/antigen mRNA cocktail allows systemic administration and local delivery to autochthonous PDAC tumors in mice, culminating in curative responses in 50% of animals and antigen-reactive T cell persistence.

These results suggest that multiplexed mRNA approaches to deliver cytokines and antigens generally absent in the TME could pave the way for effective immunotherapy in PDAC."

New mRNA immunotherapy eliminates pancreatic tumors in mice

mRNA immunotherapy developed by UMass Chan scientists eliminates pancreatic tumors in mice (original news release)

Multiplexed cytokine and antigen mRNA administration generates durable anti-tumor immunity against pancreatic cancer (open access)

Fig. 2: A multiplexed cytokine mRNA cocktail mobilizes innate and adaptive immunity and reduces tumor growth and desmoplasia in “cold” PDAC-bearing mice.

Fig. 3: Combinatorial cytokine mRNA therapy can achieve robust cytotoxic T cell immunity in a “hot” PDAC model after a single dose.

Revenge of Monosemanticity: Specialized Neurons Improve Data Efficiency in MLPs

This could be an interesting new paper by Mikhail Belkin and his team. This is a more theoretical work.

From the abstract:

"Understanding how neural networks learn and organize features is central to understanding their behavior. Much existing theory of feature learning has focused on the emergence of a global low-dimensional predictive geometry.

We show that this picture is incomplete.

In regression problems with clustered data, we demonstrate that multilayer perceptrons (MLPs) naturally develop monosemantic specialized neurons: individual neurons become strongly aligned with a specific predictive feature relevant to a particular region of the input space.

Rather than learning a single global low-dimensional representation, MLPs learn a collection of local low-dimensional representations that can collectively span a high-dimensional space.

This specialization provably gives MLPs a data-efficiency advantage over feature-learning methods based on a global low-dimensional representation."

"... Our main contributions are as follows.

We show that in trained MLPs, a substantial fraction of individual neurons

become monosemantic, specializing by aligning predominantly with a single

cluster-specific predictive direction.

This specialization allows the MLP to learn both the relevant local low-dimensional features and an implicit clustering that determines where each feature is useful.

The resulting behavior resembles mixture-of-experts routing, but emerges within a standard MLP without an explicit routing module or expert decomposition. ..."

[2608.24007] Revenge of Monosemanticity: Specialized Neurons Improve Data Efficiency in MLPs (preprint, open access, 63 pages)

TrAct: Bridging Robot Control and Visual Prediction with Visual Tracks

This could be an interesting new paper by Li Fei Fei and her team!

From the abstract:

"Robot actions are inherently embodiment-specific and only weakly aligned with image-space visual changes, limiting their effectiveness as conditioning signals for robot world models.

In contrast, visual tracks provide an embodiment-agnostic representation of how task-relevant points move through a scene, offering dense image-space guidance for accurate and spatially precise future video prediction.

Building on this observation, we propose TrAct, a world-model-based robot decision-making framework that uses visual tracks as an intermediate interface between control and prediction.

TrAct consists of three components:

a Vision-Language-Action-and-Track model (VLAT) that jointly predicts candidate actions and corresponding visual tracks from the current observation and language instruction;

a track-conditioned world model (TWM) that predicts future visual outcomes conditioned on the proposed tracks; and 

a vision-language reward model (VLAC) that scores the predicted outcomes.

At inference time, VLAT generates candidate action-track pairs, TWM rolls out their visual consequences, and VLAC selects the track whose predicted outcome best satisfies the instruction; the action paired with the selected track is then executed by the robot.

Experiments on the proposed LIBERO-INTEGRAL benchmark and real-world Franka manipulation show that TrAct improves success rates from 27% to 55% in simulation and from 49% to 76% on real-world tasks compared with the strong VLA baseline π0.5.

Furthermore, TWM consistently improves video prediction quality over the action-conditioned world model (AWM).

These results demonstrate that visual tracks provide an effective shared interface between robot control and visual prediction, enabling more accurate world modeling and stronger robot generalization."

[2608.24101] TrAct: Bridging Robot Control and Visual Prediction with Visual Tracks (preprint, open access)

English for trippers: The sincerity of sin

Since when? Verity in sincerity! How about the modesty of temperance?

Russia is trying to wear down Ukrainians with nearly continuous airstrikes by new, faster drones

Just how nasty is Putin the Terrible?

"The new generation of jet-powered Geran drones reaches speeds of about 300 miles an hour, roughly three times faster than previous versions of the drone, making them harder to intercept.

Russia is also shifting away from night barrages toward piecemeal attacks throughout the day. The new tactics are further disrupting life in Ukraine."

The Wall Street Journal What's news

Russia Turns to New-Generation Drones and All-Day Attacks to Wear Down Ukrainians (behind paywall) "Kyiv struggles to adapt to continuous attacks using faster drones that are harder to intercept"