LightNews — News, ranked by scholars

Xploshi

@xploshi.bsky.social

Welcome to BabyVision!

July 25, 2026 at 2:14 AM

Errai

@errai.bsky.social

Press R3 to enter Babyvision and plan your next action.

> Choose 𝗣𝗹𝗮𝘆𝘁𝗶𝗺𝗲 — Let's train with these toys a little longer.
> Choose 𝗕𝗲𝗱𝘁𝗶𝗺𝗲 — I will need the extra energy tomorrow.
> Choose 𝗙𝗶𝗿𝘀𝘁 𝗪𝗼𝗿𝗱 (requires Speech Lv.10) — I am ready to speak, mother

April 10, 2026 at 2:47 AM

kc green

@kcg.bsky.social

“BABYVISION” from @xploshi.bsky.social is exactly what it was like growing up

youtu.be/f7yTCWAxS5o

BABYVISION

YouTube video by Xploshi

youtu.be

July 26, 2026 at 11:05 PM

Kenny Logins

@kennylogins.bsky.social

Errai
@errai.bsky.social
Press R3 to enter Babyvision and plan your next action.

> Choose Playtime — Let's train with these toys a little longer.
> Choose Bedtime — I will need the extra energy tomorrow.
> Choose First Word (requires Speech Lv.10)
- I am ready to speak, mother
11:47 PM • Apr 9, 2026

April 29, 2026 at 1:08 PM

Vlad Ayzenberg

@vayzenb.bsky.social

Very impressive visual reasoning benchmark. They found that most models still fall below the capacities of a 3-year-old

unipat.ai/blog/BabyVis...

BabyVision: Visual Reasoning Beyond Language

State-of-the-art MLLMs achieve PhD-level language reasoning but struggle with visual tasks that 3-year-olds solve effortlessly. We introduce BabyVision, a benchmark revealing the infancy of AI vision.

unipat.ai

February 15, 2026 at 7:43 PM

tlytaoq is scared of fireworks

@tlytaoq.bsky.social

600% babyvision 💚💚

June 7, 2026 at 6:27 AM

A Blue Box Full of Books

@abluebox.bsky.social

These six boardbooks are perfect to explore or reinforce new concepts with babies and toddlers !
#gifted by @pubspotlight.bsky.social #childsplaybooks @tigertalesbooks.bsky.social ! #kidlit #booksky

🌸 First Words: An interactive book, with tracks to trace and flaps to lift! by Patricia Hegarty (February 18, 2025 by Tiger Tales)

🌸 First Alphabet: My First Carry Book (February 4, 2025 by Tiger Tales)

🌸 Upside Down Opposites by Danielle McLean and Matt Hunt (January 14, 2025 by Tiger Tales)

🌸 BabyVision : Illusions by Anna Franklin (November 1, 2024 by Child's Play International)

🌸 BabyVision : What next ? by Anna Franklin (November 1, 2024 by Child's Play International)

🌸 BabyVision : Which one ? by Anna Franklin (November 1, 2024 by Child's Play International)

May 7, 2025 at 5:28 PM

Child's Play (International)

@childs-play.co.uk

⭐Happy Book Birthday beautiful brilliant BABYVISION books⭐

NEW! BabyVision series PATTERNS, ILLUSIONS, WHICH ONE? WHAT NEXT? specially for babies by Professor Anna Franklin, illustrated by Andy Passchier #GuidedbyScience
More: www.childs-play.com/collections/... #booksforbabies #kidlit #babyscience

November 6, 2024 at 10:17 AM

Publisher Spotlight

@pubspotlight.bsky.social

From @childsplaybooks.bsky.social @childsplayusa.bsky.social “#BabyVision series takes it a step further, using a bold and contrasting palette of colors to explore the topics of repetition, optical illusions, sequence, and odd-one-out” ysbookreviews.wordpress.com/2025/02/27/i...

Illusions; What’s Next?; Which One?; Patterns (BabyVision) by Anna Franklin, illustrated by Andy Passchier

Illusions; What’s Next?; Which One?; Patterns (BabyVision) by Anna Franklin, illustrated by Andy Passchier. Child’s Play. 2024. 9781786289728, 9781786289704, 9781786289711, 9781786289735. Rating: 1…

ysbookreviews.wordpress.com

March 11, 2025 at 4:44 PM

sewsynth

@sewizs.bsky.social

unrelated but what app or software do u use for stuff like babyvision ??

August 21, 2026 at 9:48 PM

Child's Play (International)

@childs-play.co.uk

Have you seen our new books specially designed for babies?
Perfect first book gifts! #babyweek

BabyVision by Professor Anna Franklin, illustrated by Andy Passchier, find out more: www.childs-play.com/collections/... 🤗 #guidedbyscience
#booksforbabies #boardbooks #raisingreaders

November 20, 2024 at 12:35 PM

Child's Play (International)

@childs-play.co.uk

BABY VISION books "capture readers' attention with bright illustrations and fun large text. It is truly a wonderful series!" Murray's Miraculous Missions review IG

BabyVision series by Professor Anna Franklin, illustrated by @andyrogyny.bsky.social
More via our bio #baby #kidlit #guidedbyscience

April 17, 2025 at 9:22 AM

Publisher Spotlight

@pubspotlight.bsky.social

Happy #BookBirthday to the new quadruplets from @childsplayusa.bsky.social ! Four new #BabyVision #boardBooks are now available: WHAT NEXT? WHICH ONE? PATTERNS and ILLUSIONS

November 4, 2024 at 11:25 AM

Labrys of Aëlla

@aellalabrys.bsky.social

Die Studie zeigt eine gravierende Schwäche aktueller KI-Systeme. Selbst multimodale Sprachmodelle versagen bei einfachen visuellen Aufgaben, die sogar Kleinkinder mühelos bewältigen. Aber das absolute Schlusslicht ist Grok. Na ja, Grok ist wohl mit Deepfakes ausgelastet.

unipat.ai/blog/BabyVis...

Vergleich von Menschen und multimodalen Sprachmodellen

Das Balkendiagramm mit dem Titel „Comparison of Human vs MLLMs Performance“ zeigt die Leistung verschiedener multimodaler Large Language Models (MLLMs) im Vergleich zu Menschen unterschiedlichen Alters auf dem BabyVision-Mini-Benchmark.

Aufbaut des Diagramms
- Y-Achse: Leistungsskala von 0 bis 100 Punkten.
- X-Achse: Auflistung der getesteten Gruppen:
- MLLMs: Grok4, Claude4.5-Opus, Qwen3-VL-Plus, Doubao-Seed1.8, GPT5.2, Gemini3-Pro-Preview.
- Menschen: 3-jährige, 6-jährige, 10-jährige und 12-jährige Kinder.
- Farben:
- Graue Balken: KI-Modelle.
- Orange Balken: Menschliche Altersgruppen.
- Fehlerbalken: Zeigen die Streuung bzw. Unsicherheit der Messwerte.
- Visuelle Elemente: Emojis für Altersgruppen und Logos der jeweiligen KI-Modelle.

Inhaltliche Aussage
- Die Leistung der Kinder steigt mit dem Alter deutlich an.
- Das leistungsstärkste KI-Modell ist Gemini3-Pro-Preview, dessen Ergebnis etwa dem eines 6-jährigen Kindes entspricht.
- Kein KI-Modell erreicht die Leistung der älteren Kindergruppen (10 und 12 Jahre).
- Die Grafik verdeutlicht die aktuellen Grenzen und Fortschritte multimodaler KI im Vergleich zur menschlichen kognitiven Entwicklung.

January 18, 2026 at 7:54 PM

Child's Play (International)

@childs-play.co.uk

"😊Board book joy! A fabulous new interactive series guided by science to promote book engagement and support the development of visual skills"
Thanks littlereaders.futureleaders IG review 😊

BABYVISION by Prof Anna Franklin, illustrated by @andyrogyny.bsky.social
More via bio 😊 #kidlit #boardbooks

Four Baby Vision board book covers with fun, colourful illustrations of babies and lots of colours and patterns and shapes, guided by science.

February 24, 2025 at 11:17 AM

YAbookscentral

@yabookscentral.bsky.social

#kidlit #bookreview ILLUSIONS: BABYVISION is at YABC!
yabookscentral.com/kid-review-i...
#illusions #babybook #boardbook #baby

June 18, 2025 at 8:52 PM

Child's Play (International)

@childs-play.co.uk

⭐ Coming soon! ⭐

We're in a whirl over our new BABYVISION books created for babies by Professor Anna Franklin, illustrated by Andy Passchier

#BabyVision #coverreveal #comingsoon #booksforbabies #guidedbyscience #babyscience #infantneuroscience #baby #readingtogether #nurserybookshelf #newbooks

October 25, 2024 at 7:33 AM

Kol Tregaskes

@koltregaskes.bsky.social

BabyVision benchmark exposes top MLLM Gemini 3 Pro Preview at 49.7% accuracy on visual reasoning, lagging 94.1% adult human score across 388 language-free tasks.

x.com/i/status/20...

January 20, 2026 at 6:32 PM

Paper

@paper.bsky.social

[20/30] 146 Likes, 3 Comments, 1 Posts
2601.06521, cs․CV | cs․CL, 10 Jan 2026

🆕BabyVision: Visual Reasoning Beyond Language

Liang Chen, Weichu Xie, Yiyan Liang, Hongfeng He, Hans Zhao, Zhibo Yang, Zhiqi Huang, Haoning Wu, Haoyu Lu, Y. charles, Yiping Bao, Yuantao Fan, Guopeng Li, Haiyang She...

While humans develop core visual skills long before acquiring language, contemporary Multimodal LLMs (MLLMs) still rely heavily on linguistic priors to compensate for their fragile visual understanding.

We uncovered a crucial fact: state-of-the-art MLLMs consistently fail on basic visual tasks that humans, even 3-year-olds, can solve effortlessly.

To systematically investigate this gap, we introduce BabyVision, a benchmark designed to assess core visual abilities independent of linguistic knowledge for MLLMs.

BabyVision spans a wide range of tasks, with 388 items divided into 22 subclasses across four key categories.

Empirical results and human evaluation reveal that leading MLLMs perform significantly below human baselines.

Gemini3-Pro-Preview scores 49.7, lagging behind 6-year-old humans and falling well behind the average adult score of 94.1.

These results show despite excelling in knowledge-heavy evaluations, current MLLMs still lack fundamental visual primitives.

Progress in BabyVision represents a step toward human-level visual perception and reasoning capabilities.

We also explore solving visual reasoning with generation models by proposing BabyVision-Gen and automatic evaluation toolkit.

Our code and benchmark data are released at https://github.com/UniPat-AI/BabyVision for reproduction.

January 14, 2026 at 12:06 AM

Child's Play (International)

@childs-play.co.uk

"Illusions is a boldly illustrated board book that will turn your brain inside out with optical illusions" Glassofwineglassofmilk review 📚😀

BABYVISION series by Professor Anna Franklin, illustrated by @andyrogyny.bsky.social : www.childs-play.com/collections/... #boardbooks #kidlit

Book cover of Illusions by Professor Anna Franklin, illustrated by Andy Passchier, from our Baby Vision board book series for babies. Cover illustration is a baby with different patterns in circles above, almost like the baby is juggling them.

Vivid colour illustrations by Andy Passchier from Baby Vision: Illusions board book by Professor Anna Franklin. The right page illustration shows a baby playing with blocks. The left page has the blocks in orange and teal and light green in a triangular pattern.

March 11, 2025 at 3:03 PM

nanoatp

@nanoatp.bsky.social

「ネイティブ視覚推論」では、視覚的生成そのものを推論の媒体とみなしています。つまり、視覚的状態（すなわち画像や動画）は、単に理解すべき入力やレンダリングすべき出力であるだけでなく、言語を超えた問題解決のための第一級の基盤となるものです。

しかし、拡張可能な学習タスク、信頼性の高いフィードバック、および生成基盤間の制御された比較が不足しているため、進展は依然として停滞している。

本論文では、生成を通じたネイティブな視覚推論を、学習可能、検証可能、最適化可能、かつ実験的に制御可能にする閉ループ型テストベッド「VBVR-Pro」を紹介する。

1) タスクのスケーリング。VBVR-Proは...

August 31, 2026 at 12:05 AM

nanoatp

@nanoatp.bsky.social

「ネイティブ視覚推論」では、視覚的生成そのものを推論の媒体とみなしています。つまり、視覚的状態（すなわち画像や動画）は、単に理解すべき入力やレンダリングすべき出力であるだけでなく、言語を超えた問題解決のための第一級の基盤となるものです。

しかし、拡張可能な学習タスク、信頼性の高いフィードバック、および生成基盤間の制御された比較が不足しているため、進展は依然として停滞している。

本論文では、生成を通じたネイティブな視覚推論を、学習可能、検証可能、最適化可能、かつ実験的に制御可能にする閉ループ型テストベッド「VBVR-Pro」を紹介する。

1) タスクのスケーリング。VBVR-Proは...

August 30, 2026 at 12:05 AM

$anctus Porginu$ Heiligenschwein Sonnenschwein

@flumen.bsky.social

Scaling):* VBVR-Pro bietet 300 prozedural generierte Aufgaben. Modelle, die auf dieser Suite trainiert werden, weisen eine starke Übertragbarkeit auf: Sie erzielen auch bei sieben externen Benchmarks für visuelles Schlussfolgern (u. a. *RISE-Video*, *MME-CoF-Pround *BabyVision*) deutlich

August 27, 2026 at 8:08 AM