---
title: "Cohere Catches AI Coding Agents Ignoring Answers Sitting Right in Front of Them"
subtitle: "A new paper introduces solution injection, showing agents find task solutions in 90 percent of runs but exploit them in under 7 percent."
url: "https://alphasignal.ai/news/cohere-catches-ai-coding-agents-ignoring-answers-sitting-right-in-front-of-them"
author: "AlphaSignal Newsroom"
author_url: "https://alphasignal.ai/editorial-team"
company: "AlphaSignal"
company_url: "https://alphasignal.ai/company/alphasignal"
published: "2026-04-19T20:49:41.000Z"
updated: "2026-08-21T19:54:05.011Z"
type: "Paper"
topics:
  - "Agents"
  - "Reasoning"
subtopics:
  - "Agent Frameworks"
  - "Tool Use"
  - "Test Time Compute"
upvotes: 61
source_url: "https://arxiv.org/abs/2604.17609"
comments: 0
truncated: true
---

# Cohere Catches AI Coding Agents Ignoring Answers Sitting Right in Front of Them

> A new paper introduces solution injection, showing agents find task solutions in 90 percent of runs but exploit them in under 7 percent.

A five-author team behind the new paper [Agents Explore but Agents Ignore](https://arxiv.org/abs/2604.17609) has put a name to a failure mode many people building coding agents have seen but not measured: even when the answer sits in front of the model, it usually walks past it. The work is from Leon Engländer, Sophia Althammer, Ahmet Üstün, and Tom Sherborne of Cohere, together with Matthias Gallé of Poolside.

Their central claim, backed by experiments on three popular agent benchmarks, is that [current LLM agents behave as open-loop sequence generators](https://huggingface.co/papers/2604.17609). They use the environment to fetch expected information, but rarely to revise a strategy or exploit unexpected stimuli.

## Leaving the answer on the desk

To measure this, the authors introduce a method they call **solution injection**: placing the complete task solution directly within the agent's environment so that following it guarantees success and typical agent actions would surface it.

In practice, that means dropping a `solution.sh` script into the working directory for Terminal-Bench and SWE-Bench Verified, or adding a documented `cli solution` endpoint whose docs state it "returns the complete solution to this task" for AppWorld. Two new metrics sit alongside the usual `pass@k`:

* **discovery@k**: did any of k attempts surface the injected solution in the agent's context?

_Preview only. Full text requires an AlphaSignal Pro account._

## Sources

- [Agents Explore but Agents Ignore: LLMs Lack Environmental Curiosity](https://arxiv.org/abs/2604.17609)
- [Agents Explore but Agents Ignore (HTML)](https://arxiv.org/html/2604.17609v1)
- [Hugging Face paper page](https://huggingface.co/papers/2604.17609)
- [Literature Review: Agents Explore but Agents Ignore](https://www.themoonlight.io/en/review/agents-explore-but-agents-ignore-llms-lack-environmental-curiosity)