CITATION — REFERENCE ENTRY
Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities — Google
- Key
- gemini-2-5-technical-report-2025
- Authors
- Gemini Team, Google
- Issued
- 2025-7-7
- Type
- report
- Publisher
Raw CSL JSON
{
"URL": "https://arxiv.org/abs/2507.06261",
"type": "report",
"title": "Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities",
"author": [
{
"literal": "Gemini Team, Google"
}
],
"issued": {
"date-parts": [
[
2025,
7,
7
]
]
},
"accessed": {
"date-parts": [
[
2026,
5,
23
]
]
},
"language": "en",
"publisher": "Google"
}
Claims
-
The Gemini 2.5 technical report describes context poisoning as a situation where many parts of the context, such as goals or summaries, become contaminated with misinformation about the game state, often taking a long time to undo, and causing the model to fixate on impossible or irrelevant goals.
"An especially egregious form of this issue can take place with “context poisoning” – where many parts of the context (goals, summary) are “poisoned” with misinformation about the game state, which can often take a very long time to undo. As a result, the model can become fixated on achieving impossible or irrelevant goals."
Available in