Skip to content
OpenTrain AIFor AI Companies

HFEPX · Eval paper review

Prompt-Induced Waste in Coding Agents: Reasoning, Effort, Harness Design, and End-to-End Cost

Sarel Weinberger, Amir Hozez

Published

Aug 2, 2026

Citations

0

Trust level

Moderate

Usefulness score

25/100 (Low)

Extraction confidence

55% (Moderate)

Derived from extracted protocol signals and abstract evidence.

Rater population

Not reported

Signals refreshed

Aug 21, 2026

Open arXiv Find implementation Shortlist (0) RSS feed

Tags

Automatic Metrics Coding Long Horizon Tool Use

Should you rely on this paper?

This paper is adjacent to HFEPX scope and is best used for background context, not as a primary protocol reference.

Use this for comparison and orientation, not as your only source.

Best use

Background context only

Use if you need

A benchmark-and-metrics comparison anchor.

What to verify

Validate the evaluation procedure and quality controls in the full paper before operational use.

Main weakness

No major weakness surfaced.

Human feedback signal
Not explicit
Not explicit in abstract metadata
Evaluation signal
Detected
Eval setup described
Usefulness for eval research
25/100
Adjacent candidate

Treat as adjacent context, not a core eval-method reference.

If you are doing eval pipeline work, start here

Human Eval Papers LLM-as-Judge Papers Pairwise Preference Papers Tool-Use Evaluation Search

Abstract

Coding-agent efficiency cannot be characterized by token count or model price alone. We study how end-to-end cost and task success depend jointly on prompt semantics, inference effort, harness policy, model, task difficulty, tool use, context management, and provider accounting. Controlled prompt experiments show that wording can change reasoning and verification behavior without changing the task. A separate SWE-bench Verified study shows that additional inference effort can improve difficult tasks for some models but can also add cost without benefit. A DeepSeek Harness extension shows that the effect of an effort-control intervention changes substantially when the harness changes, even when the model, tasks, prompts, and controller logic are held fixed. These results show that prompt, effort, and harness are interacting experimental factors rather than independent efficiency controls. We model efficiency as cost per successful task induced by the agent trajectory. Token and cache counts are measurements of that trajectory, not sufficient optimization targets. Agent evaluations should therefore measure success and end-to-end cost while controlling the system variables that determine how the trajectory is produced

What we could verify

These are the protocol signals we could actually recover from the available paper metadata. Use them to decide whether this paper is worth deeper reading.

Human Feedback Types

missing

None explicit

No explicit feedback protocol extracted.

"Coding-agent efficiency cannot be characterized by token count or model price alone."

Evaluation Modes

strong

Automatic Metrics

Includes extracted eval setup.

"Coding-agent efficiency cannot be characterized by token count or model price alone."

Quality Controls

missing

Not reported

No explicit QC controls found.

"Coding-agent efficiency cannot be characterized by token count or model price alone."

Benchmarks / Datasets

strong

SWE Bench, SWE Bench Verified

Useful for quick benchmark comparison.

"A separate SWE-bench Verified study shows that additional inference effort can improve difficult tasks for some models but can also add cost without benefit."

Reported Metrics

strong

Task success

Useful for evaluation criteria comparison.

"We study how end-to-end cost and task success depend jointly on prompt semantics, inference effort, harness policy, model, task difficulty, tool use, context management, and provider accounting."

Benchmarks and datasets

SWE-benchSWE-bench Verified

Reported metrics

task success
Human feedback details
Uses human feedback
No
Feedback types
None
Rater population
Not reported
Unit of annotation
Trajectory
Expertise required
Coding
Evaluation details
Evaluation modes
Automatic Metrics
Agentic eval
Tool Use, Long Horizon
Quality controls
Not reported
Evidence quality
Moderate
Use this page as
Background context only

Research brief

Metadata summary

Coding-agent efficiency cannot be characterized by token count or model price alone.

Based on abstract + metadata only. Check the source paper before making high-confidence protocol decisions.

Key takeaways

  • Coding-agent efficiency cannot be characterized by token count or model price alone.
  • We study how end-to-end cost and task success depend jointly on prompt semantics, inference effort, harness policy, model, task difficulty, tool use, context management, and provider accounting.
  • Controlled prompt experiments show that wording can change reasoning and verification behavior without changing the task.

Researcher actions

  • Compare this paper against others mentioning SWE-bench.
  • Validate inferred eval signals (Tool-use evaluation) against the full paper.
  • Use related-paper links to find stronger protocol-specific references.

Caveats

  • Generated from abstract + metadata only; no PDF parsing.
  • Signals below are heuristic and may miss details reported outside the abstract.

Recommended queries

SWE-bench

Contribution summary

  • Coding-agent efficiency cannot be characterized by token count or model price alone.
  • We model efficiency as cost per successful task induced by the agent trajectory.
  • Agent evaluations should therefore measure success and end-to-end cost while controlling the system variables that determine how the trajectory is produced

Why it matters for eval

  • Coding-agent efficiency cannot be characterized by token count or model price alone.
  • We model efficiency as cost per successful task induced by the agent trajectory.

Researcher checklist

  • Human feedback protocol is explicit

    No explicit human feedback protocol detected.

  • Evaluation mode is explicit

    Detected: Automatic Metrics

  • Quality control reporting appears

    No calibration/adjudication/IAA control explicitly detected.

  • Benchmark or dataset anchors are present

    Detected: SWE-bench, SWE-bench Verified

  • Metric reporting is present

    Detected: task success