Get the newsletter
Salesforce · AI / ML · Filed Jul 31, 2025 · USPTO published Aug 20, 2026 · Aug 20 release

Salesforce Patents AI That Corrects Its Own Visual Math Mistakes

Counting objects in a photo sounds trivial for a human, but it trips up most AI systems. Salesforce is filing a patent for a method that trains AI to get this right by having one model quiz another, then scoring the answers.

Visual reasoning task examples including volume, distance, length, slope, and angle comparisons. Drawing from patent filing US 2026/0244680 A1.
Visual reasoning task examples including volume, distance, length, slope, and angle comparisons.All 13 drawings ↓
Publication number
US 2026/0244680 A1(US20260244680A1)
Applicant
Salesforce, Inc.
Filed · Published
Jul 31, 2025 · Aug 20, 2026
Inventors
Kung-Hsiang Huang, Can Qin, Shafiq Rayhan Joty, Chien-Sheng Wu
Status when we published
Examiner raised objections; the company can respond
Document
20 claims
US classification
707/769
Examiner
HU, JENSEN (Art Unit 2169)
Parent application
Claims priority from a provisional application 63758853 (filed Feb 14, 2025)

How Salesforce trains AI to count and compare what it sees

You're reviewing a sales dashboard packed with charts and product photos, and the AI assistant you rely on confidently gives you the wrong count. It says there are four items in a cluster image when there are six. That kind of arithmetic slip is a real and common failure in AI systems that read images.

Salesforce's patent describes a system where AI teaches itself to fix that problem. A first AI model generates an image, writes a math-related question about it (like "how many red objects are in this picture?"), and produces both a correct and a deliberately wrong answer. That set of four pieces, the image, the question, the right answer, and the wrong one, is used to train a second AI model. The second model learns by figuring out which answer is right and updating itself based on whether it got it correct.

The result is an AI agent that has practiced thousands of these drills before ever seeing real data. Rather than needing humans to label millions of examples by hand, the system generates its own training material automatically.

From the filing · CLAIM 1
generating, programmically, an image containing at least one object having a target arithmetic property; generating a query relating to the target arithmetic property using a query-response template based on the generated image …

Translation: The system creates an image with a specific math feature and makes up a matching question to test the AI.

Inside the four-part training data Salesforce builds from scratch

The patent describes a pipeline for generating training data and using it to improve a language model's ability to reason about numbers it sees in images, things like counting objects, comparing quantities, or identifying numeric properties in a picture.

Here is how the pipeline works step by step:

  • Step 1 - Synthetic image generation: A program creates images that contain objects with specific arithmetic properties (for example, seven blue circles or three stacked boxes). This removes the need for human-labeled photo datasets.
  • Step 2 - Query and answer generation: A first neural network (a large language model) reads the image and writes a question tied to the numeric property, then generates both a correct "positive" response and an incorrect "negative" response. Together with the image and question, these form what the patent calls a training quadruple.
  • Step 3 - Policy-guided candidate generation: A second, separate language model is given the quadruple and produces its own candidate answer. It also outputs two probabilities: how likely is this answer to be the correct one, and how likely is it to be the wrong one.
  • Step 4 - Training objective: The system calculates a loss (a math score measuring how far off the model's guess is) based on those two probabilities and uses it to update the second model's internal settings, nudging it toward better answers over many rounds.

Once training is complete, the second model is deployed as the active AI agent. The whole loop is self-contained: no human labelers are needed at any point.

From the filing · THE ABSTRACT
… generating, by a first neural network language model, a positive response and a negative response; and forming a training quadruple including the image, the query, the positive response and the negative response.

Translation: One AI model generates both a correct and an incorrect answer to package into a training set alongside the image and question.

What this means for AI agents handling real business data

For anyone building or buying AI tools that read charts, invoices, product images, or any document where numbers matter, this kind of failure, a confident but wrong count, is not a minor annoyance. It is a trust-destroying bug. A system trained the way Salesforce describes here would have seen a huge volume of arithmetic drills before deployment, which should reduce that class of error meaningfully.

Salesforce's customer base runs on data-heavy workflows: sales reports, inventory counts, financial summaries. An AI agent that can reliably read a chart and tell you the right number has obvious value in that context. The self-generating training approach also matters economically because human data labeling is slow and expensive. AI researchers who follow new Big Tech patents in the visual reasoning space will recognize this filing as part of a wider push to close the gap between what language models can say and what they can actually count.

Salesforce files its ninth patent we've tracked since July in our AI teams watchlist, building on a task-routing system and an AI answer coach to push further into multi-agent design.

Source: US 2026/0244680 A1 at the USPTO (PDF). Full patent text and figures are in the official publication.

Get our take in your Top Stories

Liked this breakdown? Add Patentlyze as a preferred source on Google, and our plain-English take shows up more often in your Top Stories the next time Salesforce patent news breaks.

There are more where this came from

We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.

The drawings · 13 sheets

13 drawing sheets from US 2026/0244680 A1 · click any drawing to enlarge

User interacting with AI agent, drawing sheet 1 from US 2026/0244680 A1
Sheet 1 User interacting with AI agent
Automated image generation and training pipeline, drawing sheet 2 from US 2026/0244680 A1
Sheet 2 Automated image generation and training pipeline
Visual reasoning and measurement tasks, drawing sheet 3 from US 2026/0244680 A1
Sheet 3 Visual reasoning and measurement tasks
Training dataset processing and model interaction, drawing sheet 4 from US 2026/0244680 A1
Sheet 4 Training dataset processing and model interaction
Computing device architecture and memory modules, drawing sheet 5 from US 2026/0244680 A1
Sheet 5 Computing device architecture and memory modules
Neural network layers and connection weights, drawing sheet 6 from US 2026/0244680 A1
Sheet 6 Neural network layers and connection weights
Connected server and database network infrastructure, drawing sheet 7 from US 2026/0244680 A1
Sheet 7 Connected server and database network infrastructure
Training method flowchart for neural models, drawing sheet 8 from US 2026/0244680 A1
Sheet 8 Training method flowchart for neural models
Show the other 5 sheets
Salesforce Patents AI That Corrects Its Own Visual Math Mistakes patent drawing, sheet 9
Sheet 9
Salesforce Patents AI That Corrects Its Own Visual Math Mistakes patent drawing, sheet 10
Sheet 10
Salesforce Patents AI That Corrects Its Own Visual Math Mistakes patent drawing, sheet 11
Sheet 11
Salesforce Patents AI That Corrects Its Own Visual Math Mistakes patent drawing, sheet 12
Sheet 12
Salesforce Patents AI That Corrects Its Own Visual Math Mistakes patent drawing, sheet 13
Sheet 13
Patent filing page

More from Salesforce

All 110 Salesforce patents →

Watchlists this filing belongs to

All watchlists →