Skip to content
← Back to explorer

HFEPX Hub

Long Horizon + General Papers

Updated from current HFEPX corpus (Apr 12, 2026). 114 papers are grouped in this hub page.

Read Full Context

Updated from current HFEPX corpus (Apr 12, 2026). 114 papers are grouped in this hub page. Common evaluation modes: Automatic Metrics, Simulation Env. Most common rater population: Domain Experts. Common annotation unit: Trajectory. Frequent quality control: Calibration. Frequently cited benchmark: ALFWorld. 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: 114 Last published: Mar 22, 2026 Global RSS Tag RSS
Long HorizonGeneral

Researcher Quick Triage

This hub is best used for protocol triage and replication planning from abstract-level evidence. Quality band: High .

Analysis blocks below are computed from the currently loaded sample (60 of 114 total papers in this hub).

High-Signal Coverage

100.0%

60 / 60 sampled papers are not low-signal flagged.

Replication-Ready Set

17

Benchmark + metric + eval mode explicitly present.

Judge/Human Comparability

1

Papers containing both `human_eval` and `llm_as_judge`.

  • 17 papers are replication-ready (benchmark + metric + explicit evaluation mode).
  • 1 papers support judge-vs-human agreement analysis.
  • 1 papers report explicit quality controls (calibration/adjudication/IAA).

Primary action: Start with the top 2 papers in “Start Here”, then validate assumptions in the protocol matrix.

Currently showing only replication-ready papers in ranking and matrix sections (17 papers).

Need evaluators for this research workflow?

Post a Job →

Why This Matters For Eval Research

  • 25.4% of papers report explicit human-feedback signals, led by pairwise preferences.
  • automatic metrics appears in 63.2% of papers in this hub.
  • ALFWorld is a recurring benchmark anchor for cross-paper comparisons in this page.

Protocol Takeaways

  • 1 sampled papers report both human evaluation and LLM-as-judge, supporting direct agreement checks.
  • Most common quality-control signal is rater calibration (0.9% of papers).
  • Rater context is mostly domain experts, and annotation is commonly trajectory-level annotation; use this to scope replication staffing.

Benchmark Interpretation

  • ALFWorld appears in 2.6% of hub papers (3/114); use this cohort for benchmark-matched comparisons.
  • BrowseComp appears in 2.6% of hub papers (3/114); use this cohort for benchmark-matched comparisons.

Metric Interpretation

  • accuracy is reported in 29.8% of hub papers (34/114); compare with a secondary metric before ranking methods.
  • cost is reported in 17.5% of hub papers (20/114); compare with a secondary metric before ranking methods.
Researcher Checklist (Expanded)

Researcher Checklist

  • Gap: Papers with explicit human feedback

    Coverage is a replication risk (25.4% vs 45% target).

  • Gap: Papers reporting quality controls

    Coverage is a replication risk (0.9% vs 30% target).

  • Moderate: Papers naming benchmarks/datasets

    Coverage is usable but incomplete (29.8% vs 35% target).

  • Strong: Papers naming evaluation metrics

    Coverage is strong (71.9% vs 35% target).

  • Gap: Papers with known rater population

    Coverage is a replication risk (6.1% vs 35% target).

  • Strong: Papers with known annotation unit

    Coverage is strong (47.4% vs 35% target).

Strengths

  • 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 0.9% of papers report quality controls; prioritize calibration/adjudication evidence.
  • Rater population is under-specified (6.1% 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 (ALFWorld vs BrowseComp) before comparing methods.
  • Track metric sensitivity by reporting both accuracy and cost.
  • Add inter-annotator agreement checks when reproducing these protocols.
Recommended Queries (Expanded)

Recommended Queries

Start with These 3

Use these when you need one protocol anchor, one benchmark anchor, and one recent comparison point before reading the wider hub.

Start Here (Best First 6)

Ranked for protocol completeness (human signal, benchmark + metric anchors, quality controls, and judge/human overlap).

Protocol Matrix (Top 12)

Use this to quickly compare protocol ingredients instead of scanning long prose.

Paper HF Signal Eval Modes Benchmarks Metrics QC
AgentHER: Hindsight Experience Replay for LLM Agent Trajectory Relabeling

Mar 22, 2026

Yes Human Eval , Llm As Judge WebArena , ToolBench Precision , Pass@1 Not Reported
TraceSafe: A Systematic Assessment of LLM Guardrails on Multi-Step Tool-Calling Trajectories

Apr 8, 2026

Yes Automatic Metrics Tracesafe Bench Accuracy Not Reported
PEARL: Self-Evolving Assistant for Time Management with Reinforcement Learning

Jan 17, 2026

Yes Automatic Metrics Calconflictbench Error rate Not Reported
ReDAct: Uncertainty-Aware Deferral for LLM Agents

Apr 8, 2026

No
Not Reported
Simulation Env ALFWorld Cost , Token cost Not Reported
DataSTORM: Deep Research on Large-Scale Databases using Exploratory Data Analysis and Data Storytelling

Apr 7, 2026

No
Not Reported
Human Eval Insightbench Recall Not Reported
MemMachine: A Ground-Truth-Preserving Memory System for Personalized AI Agents

Apr 6, 2026

No
Not Reported
Automatic Metrics HotpotQA Accuracy , Recall Not Reported
OSCAR: Orchestrated Self-verification and Cross-path Refinement

Apr 2, 2026

No
Not Reported
Automatic Metrics RAGTruth , HotpotQA Accuracy Not Reported
Asymmetric Actor-Critic for Multi-turn LLM Agents

Mar 31, 2026

No
Not Reported
Automatic Metrics Userbench Task success Not Reported
EnterpriseLab: A Full-Stack Platform for developing and deploying agents in Enterprises

Mar 23, 2026

No
Not Reported
Automatic Metrics Enterprisearena , Enterprisebench Cost Not Reported
PASK: Toward Intent-Aware Proactive Agents with Long-Term Memory

Apr 9, 2026

No
Not Reported
Automatic Metrics Latentneeds Bench Precision , Latency Not Reported
Full-Duplex-Bench-v3: Benchmarking Tool Use for Full-Duplex Voice Agents Under Real-World Disfluency

Apr 6, 2026

No
Not Reported
Automatic Metrics Full Duplex Bench Accuracy , Pass@1 Not Reported
$\texttt{YC-Bench}$: Benchmarking AI Agents for Long-Term Planning and Consistent Execution

Apr 1, 2026

No
Not Reported
Automatic Metrics Yc Bench Cost , Inference cost Not Reported

Protocol Diff (Top Papers)

Fast side-by-side comparison for the highest-ranked papers in this hub.

Signal AgentHER: Hindsight Experience Replay for LLM Agent… TraceSafe: A Systematic Assessment of LLM Guardrail… PEARL: Self-Evolving Assistant for Time Management…
Human Feedback DemonstrationsRed TeamPairwise Preference
Evaluation Modes Human Eval, Llm As JudgeAutomatic MetricsAutomatic Metrics
Benchmarks WebArena, ToolBenchTracesafe BenchCalconflictbench
Metrics Precision, Pass@1AccuracyError rate
Quality Controls Not reportedNot reportedNot reported
Rater Population UnknownUnknownUnknown
Annotation Unit TrajectoryTrajectoryRanking
Suggested Reading Order (Extended)

This section is intentionally expanded only when needed; use “Start Here” above for a faster pass.

Suggested Reading Order

  1. Self-Debias: Self-correcting for Debiasing Large Language Models

    Start here for detailed protocol reporting and quality-control evidence. Signals: pairwise preferences. Abstract: Unlike standard preference optimization which applies broad penalties, Self-Debias employs a fine-grained trajectory-level objective subject.

  2. AgentHER: Hindsight Experience Replay for LLM Agent Trajectory Relabeling

    Include a human-eval paper to calibrate against judge-based evaluation settings. Signals: human evaluation + demonstration data. Focus: WebArena / precision. Abstract: AgentHER realises this idea through a four-stage.

  3. Beyond Rows to Reasoning: Agentic Retrieval for Multimodal Spreadsheet Understanding and Editing

    Include a human-eval paper to calibrate against judge-based evaluation settings. Signals: human evaluation. Focus: Frtr-Bench / accuracy. Abstract: Supported by over 200 hours of expert human evaluation, BRTR.

  4. Build, Judge, Optimize: A Blueprint for Continuous Improvement of Multi-Agent Consumer Assistants

    Include an LLM-as-judge paper to test judge design and agreement assumptions. Signals: LLM-as-judge + pairwise preferences. Abstract: Grocery shopping further amplifies these difficulties, as user requests are often.

  5. LifeSim: Long-Horizon User Life Simulator for Personalized Assistant Evaluation

    Adds simulation environments with pairwise preferences for broader protocol coverage within this hub. Signals: simulation environments + pairwise preferences. Focus: Lifesim-Eval. Abstract: Under both single-scenario and long-horizon settings,.

  6. TraceSafe: A Systematic Assessment of LLM Guardrails on Multi-Step Tool-Calling Trajectories

    Adds automatic metrics with red-team protocols for broader protocol coverage within this hub. Signals: automatic metrics + red-team protocols. Focus: Tracesafe-Bench / accuracy. Abstract: As large language models.

Known Limitations

Known Limitations

  • Only 0.9% of papers report quality controls; prioritize calibration/adjudication evidence.
  • Rater population is under-specified (6.1% coverage).
  • Narrative synthesis is grounded in metadata and abstracts only; full-paper implementation details are not parsed.
Research Utility Snapshot (Detailed)

Research Utility Snapshot

Human Feedback Mix

  • Pairwise Preference (18)
  • Demonstrations (7)
  • Rubric Rating (3)
  • Critique Edit (1)

Evaluation Modes

  • Automatic Metrics (72)
  • Simulation Env (29)
  • Llm As Judge (5)
  • Human Eval (3)

Top Benchmarks

  • ALFWorld (3)
  • BrowseComp (3)
  • HotpotQA (3)
  • OSWorld (3)

Top Metrics

  • Accuracy (34)
  • Cost (20)
  • Latency (9)
  • F1 (7)

Rater Population Mix

  • Domain Experts (7)

Quality Controls

  • Calibration (1)
Coverage diagnostics (sample-based): human-feedback 48.3% · benchmarks 43.3% · metrics 56.7% · quality controls 1.7%.

Top Papers

Related Hubs

Get Started

Join the #1 Platform for AI Training Talent

Where top AI builders and expert AI Trainers connect to build the future of AI.
Self-Service
Post a Job
Post your project and get a shortlist of qualified AI Trainers and Data Labelers. Hire and manage your team in the tools you already use.
Managed Service
For Large Projects
Done-for-You
We recruit, onboard, and manage a dedicated team inside your tools. End-to-end operations for large or complex projects.
For Freelancers
Join as an AI Trainer
Find AI training and data labeling projects across platforms, all in one place. One profile, one application process, more opportunities.