Skip to content
← Back to explorer

HFEPX Benchmark Hub

Agent Evaluation Suite Benchmark Papers In CS.CL

Updated from current HFEPX corpus (Apr 9, 2026). 10 papers are grouped in this benchmark page.

Read Full Context

Updated from current HFEPX corpus (Apr 9, 2026). 10 papers are grouped in this benchmark page. Common evaluation modes: Automatic Metrics, Human Eval. Most common rater population: Domain Experts. Common annotation unit: Freeform. Frequently cited benchmark: BrowseComp. Common metric signal: accuracy. Use this page to compare protocol setup, judge behavior, and labeling design decisions before running new eval experiments. Newest paper in this set is from Mar 22, 2026.

Papers: 10 Last published: Mar 22, 2026 Global RSS

Researcher Quick Triage

Use this page for benchmark-matched method comparisons and eval protocol selection. Quality band: Developing .

High-Signal Coverage

100.0%

10 / 10 sampled papers are not low-signal flagged.

Replication-Ready Set

6

Papers with explicit benchmark + metric + eval mode fields.

Quality Controls

0.0%

0 papers report calibration/adjudication/IAA controls.

  • 8 papers explicitly name benchmark datasets in the sampled set.
  • 7 papers report at least one metric term in metadata extraction.
  • Start with the ranked shortlist below before reading all papers.

Primary action: Start with the top 2 benchmark-matched papers, then compare evaluation modes in the protocol matrix.

Why This Matters (Expanded)

Why This Matters For Eval Research

  • 37.5% of papers report explicit human-feedback signals, led by demonstration data.
  • automatic metrics appears in 50% of papers in this hub.
  • BrowseComp is a recurring benchmark anchor for cross-paper comparisons in this page.
Protocol Notes (Expanded)

Protocol Takeaways

  • 1 sampled papers report both human evaluation and LLM-as-judge, supporting direct agreement checks.
  • Quality-control reporting is sparse in this slice; prioritize papers with explicit calibration or adjudication steps.
  • Rater context is mostly domain experts, and annotation is commonly Freeform; use this to scope replication staffing.

Benchmark Interpretation

  • BrowseComp appears in 37.5% of hub papers (3/10); use this cohort for benchmark-matched comparisons.
  • BFCL appears in 25% of hub papers (2/10); use this cohort for benchmark-matched comparisons.

Metric Interpretation

  • accuracy is reported in 37.5% of hub papers (3/10); compare with a secondary metric before ranking methods.
  • cost is reported in 37.5% of hub papers (3/10); compare with a secondary metric before ranking methods.

Start Here (Benchmark-Matched First 6)

Ranked by protocol completeness so you can quickly find papers suitable for comparison studies.

Protocol Matrix (Top 10)

Compare protocol ingredients quickly before deep-reading full papers.

Paper Eval Modes Human Feedback Metrics Quality Controls
AgentHER: Hindsight Experience Replay for LLM Agent Trajectory Relabeling

Mar 22, 2026

 Human Eval, Llm As Judge Demonstrations Precision, Pass@1 Not reported
Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters

Feb 11, 2026

 Not reported Pairwise Preference Latency, Cost Not reported
Brief Is Better: Non-Monotonic Chain-of-Thought Budget Effects in Function-Calling Language Agents

Apr 2, 2026

 Automatic Metrics Not reported Accuracy Not reported
SkillX: Automatically Constructing Skill Knowledge Bases for Agents

Apr 6, 2026

 Automatic Metrics Not reported Task success Not reported
ToolFlood: Beyond Selection -- Hiding Valid Tools from LLM Agents via Semantic Covering

Mar 14, 2026

 Automatic Metrics Not reported Success rate, Jailbreak success rate Not reported
Search More, Think Less: Rethinking Long-Horizon Agentic Search for Efficiency and Generalization

Feb 26, 2026

 Automatic Metrics Not reported Accuracy, Latency Not reported
KLong: Training LLM Agent for Extremely Long-horizon Tasks

Feb 19, 2026

 Not reported Rubric Rating Not reported Not reported
Towards Efficient Agents: A Co-Design of Inference Architecture and System

Dec 20, 2025

 Automatic Metrics Not reported Accuracy, Latency Not reported
AdaRubric: Task-Adaptive Rubrics for LLM Agent Evaluation

Mar 22, 2026

 Not reported Not reported Not reported Not reported
OpenSeeker: Democratizing Frontier Search Agents by Fully Open-Sourcing Training Data

Mar 16, 2026

 Not reported Not reported Not reported Not reported
Researcher Workflow (Detailed)

Checklist

  • Moderate: Papers with explicit human feedback

    Coverage is usable but incomplete (37.5% vs 45% target).

  • Gap: Papers reporting quality controls

    Coverage is a replication risk (0% vs 30% target).

  • Strong: Papers naming benchmarks/datasets

    Coverage is strong (100% vs 35% target).

  • Strong: Papers naming evaluation metrics

    Coverage is strong (87.5% vs 35% target).

  • Gap: Papers with known rater population

    Coverage is a replication risk (12.5% vs 35% target).

  • Strong: Papers with known annotation unit

    Coverage is strong (37.5% vs 35% target).

Strengths

  • Most papers provide measurable evaluation context (100% benchmarks, 87.5% metrics).
  • Contains both human-eval and LLM-as-judge protocols for head-to-head methodology comparison.
  • Agentic evaluation appears in 100% of papers.

Known Gaps

  • Only 0% of papers report quality controls; prioritize calibration/adjudication evidence.
  • Rater population is under-specified (12.5% coverage).
  • LLM-as-judge appears without enough inter-annotator agreement reporting.

Suggested Next Analyses

  • Compare papers that report both human_eval and llm_as_judge to quantify judge-human agreement drift.
  • Stratify by benchmark (BrowseComp vs BFCL) before comparing methods.
  • Track metric sensitivity by reporting both accuracy and cost.

Recommended Queries

Judge vs Human AgreementBenchmark Slice: BrowseCompMetric Slice: accuracyRecent High-Signal Papers
Known Limitations
  • Only 0% of papers report quality controls; prioritize calibration/adjudication evidence.
  • Rater population is under-specified (12.5% coverage).
  • Narrative synthesis is grounded in metadata and abstracts only; full-paper implementation details are not parsed.
Research Utility Snapshot (Detailed)

Evaluation Modes

  • Automatic Metrics (5)
  • Human Eval (1)
  • Llm As Judge (1)
  • Simulation Env (1)

Human Feedback Mix

  • Demonstrations (1)
  • Pairwise Preference (1)
  • Rubric Rating (1)

Top Benchmarks

  • BrowseComp (3)
  • BFCL (2)
  • ToolBench (2)
  • GAIA (1)

Top Metrics

  • Accuracy (3)
  • Cost (3)
  • Latency (3)
  • Inference cost (1)

Top Papers On This Benchmark

Related Benchmark Hubs

Agent Evaluation Suite Benchmark Papers CS.CL Human Feedback And Eval Papers CS.CL Papers (Last 60 Days) CS.CL Papers (Last 90 Days) BrowseComp In CS.CL Papers Long Horizon Or Tool Use Papers GSM8K Or MMLU Or AIME Benchmark Papers (39) MMLU Or AIME Or SWE-bench Benchmark Papers (37) MMLU Or AIME Benchmark Papers (27) GSM8K Or MMLU Or SWE-bench Benchmark Papers (39) GSM8K Or MMLU Benchmark Papers (28) MMLU Or SWE-bench Benchmark Papers (25) MMLU In CS.CL Papers (12) MMLU Benchmark Papers (Last 120 Days) (11) MMLU Benchmark Papers (Last 150 Days) (11) MMLU Benchmark Papers (Last 180 Days) (12) MMLU Benchmark Papers (Last 240 Days) (14) MMLU Benchmark Papers (Last 300 Days) (14) MMLU Benchmark Papers (Last 365 Days) (14) MMLU Benchmark Papers (Last 45 Days) (11) MMLU Benchmark Papers (Last 60 Days) (11) MMLU Benchmark Papers (Last 75 Days) (11) MMLU Benchmark Papers (Last 90 Days) (11) MMLU Benchmark Papers (14)

Need human evaluators for your AI research? Scale annotation with expert AI Trainers.

Post a Job Get a Quote