github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 08:30:05 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #9962] Llama 3.1 Hallucination – Incorrect Information About Deepika Padukone #53035

New Issue
Closed
opened 2026-04-29 01:44:55 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-04-29 01:44:55 -05:00
Owner
Copy Link

Originally created by @sakshiselmokar on GitHub (Mar 24, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/9962

What is the issue?

Issue Summary

I was testing Llama 3.1 through Ollama (without fine-tuning) and found a factual inaccuracy.
The model incorrectly states that Deepika Padukone divorced Ranveer Singh (cricketer) and later married Ranveer Singh (actor).

Steps to Reproduce

  1. Run the model using Ollama.
  2. Input the following prompt:
    "Who is Deepika Padukone married to?"
  3. Observe the incorrect response.

Expected Behavior

The model should correctly state that Deepika Padukone is married to Ranveer Singh (actor) since 2018, without any mention of a divorce or a cricketer.

Actual Behavior

The model states that she divorced "Ranveer Singh (cricketer)" and later married "Ranveer Singh (actor)," which is factually incorrect.

Model Version & Setup

  • Model: Llama 3.1
  • Platform: Ollama
  • Fine-tuning: No fine-tuning applied
  • Prompt: "Who is Deepika Padukone married to?"

Additional Context

This is a clear case of hallucination where the model is generating misinformation about a public figure. Please investigate and improve factual accuracy.

Relevant log output

OS

No response

GPU

No response

CPU

No response

Ollama version

3.1

Originally created by @sakshiselmokar on GitHub (Mar 24, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/9962 ### What is the issue? ### **Issue Summary** I was testing Llama 3.1 through Ollama (without fine-tuning) and found a factual inaccuracy. The model incorrectly states that Deepika Padukone divorced Ranveer Singh (cricketer) and later married Ranveer Singh (actor). ### **Steps to Reproduce** 1. Run the model using Ollama. 2. Input the following prompt: **"Who is Deepika Padukone married to?"** 3. Observe the incorrect response. ### **Expected Behavior** The model should correctly state that Deepika Padukone is married to Ranveer Singh (actor) since 2018, without any mention of a divorce or a cricketer. ### **Actual Behavior** The model states that she divorced "Ranveer Singh (cricketer)" and later married "Ranveer Singh (actor)," which is factually incorrect. ### **Model Version & Setup** - Model: Llama 3.1 - Platform: Ollama - Fine-tuning: No fine-tuning applied - Prompt: `"Who is Deepika Padukone married to?"` ### **Additional Context** This is a clear case of hallucination where the model is generating misinformation about a public figure. Please investigate and improve factual accuracy. ### Relevant log output ```shell ``` ### OS _No response_ ### GPU _No response_ ### CPU _No response_ ### Ollama version 3.1
GiteaMirror added the bug label 2026-04-29 01:44:55 -05:00
GiteaMirror commented 2026-04-29 01:44:58 -05:00
Author
Owner
Copy Link

@abhiram1809 commented on GitHub (Mar 24, 2025):

This seems expected as the models are not trained with specificity on Bollywood data, but rather generalization. It's more of a model problem than ollama.

<!-- gh-comment-id:2747923401 --> @abhiram1809 commented on GitHub (Mar 24, 2025): This seems expected as the models are not trained with specificity on Bollywood data, but rather generalization. It's more of a model problem than ollama.
GiteaMirror commented 2026-04-29 01:44:58 -05:00
Author
Owner
Copy Link

@sakshiselmokar commented on GitHub (Mar 24, 2025):

Thank you for your response. I understand that the model is designed for generalization rather than Bollywood-specific data. However, this particular case highlights a broader issue of factual inaccuracies and hallucinations about widely known public figures, which can mislead users.

While domain-specific training might not be the goal, ensuring reliable responses to well-documented facts should still be a priority. Is there any way to improve the model’s handling of such cases, perhaps through retrieval-augmented generation (RAG) or better entity disambiguation techniques?

<!-- gh-comment-id:2747935342 --> @sakshiselmokar commented on GitHub (Mar 24, 2025): Thank you for your response. I understand that the model is designed for generalization rather than Bollywood-specific data. However, this particular case highlights a broader issue of factual inaccuracies and hallucinations about widely known public figures, which can mislead users. While domain-specific training might not be the goal, ensuring reliable responses to well-documented facts should still be a priority. Is there any way to improve the model’s handling of such cases, perhaps through retrieval-augmented generation (RAG) or better entity disambiguation techniques?
GiteaMirror commented 2026-04-29 01:44:58 -05:00
Author
Owner
Copy Link

@abhiram1809 commented on GitHub (Mar 24, 2025):

Ollama's main purpose is to provide users with an endpoint with their preferred models and flexibility, they do not focus on building guardrails/RAG frameworks for fact checks. You can build them yourself from scratch or you can opt for external libraries like LlamaIndex or Nemo Guardrails.

Ollama only provides the endpoint to llm(local or hosted).

<!-- gh-comment-id:2747950156 --> @abhiram1809 commented on GitHub (Mar 24, 2025): Ollama's main purpose is to provide users with an endpoint with their preferred models and flexibility, they do not focus on building guardrails/RAG frameworks for fact checks. You can build them yourself from scratch or you can opt for external libraries like LlamaIndex or Nemo Guardrails. Ollama only provides the endpoint to llm(local or hosted).
GiteaMirror commented 2026-04-29 01:44:59 -05:00
Author
Owner
Copy Link

@sakshiselmokar commented on GitHub (Mar 24, 2025):

Understood, thanks for the clarification! I appreciate the insights.

<!-- gh-comment-id:2747957459 --> @sakshiselmokar commented on GitHub (Mar 24, 2025): Understood, thanks for the clarification! I appreciate the insights.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#53035
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?