Skip to content
← Back to explorer

HFEPX Hub

CS.CL + Automatic Metrics Papers

Updated from current HFEPX corpus (Feb 27, 2026). 812 papers are grouped in this hub page. Common evaluation modes: Automatic Metrics, Simulation Env. Most common rater population: Domain Experts. Common annotation unit: Ranking. Frequent quality control: Calibration. Frequently cited benchmark: Retrieval. 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 Feb 26, 2026.

Papers: 812 Last published: Feb 26, 2026 Global RSS Tag RSS
Cs.CLAutomatic Metrics

Research Narrative

Grounded narrative Model: deterministic-grounded Source: preview

Updated from current HFEPX corpus (Feb 27, 2026). This page tracks 812 papers for CS.CL + Automatic Metrics Papers. Dominant protocol signals include automatic metrics, simulation environments, human evaluation, with frequent benchmark focus on Retrieval, MATH and metric focus on accuracy, cost. Use the grounded sections below to prioritize reproducible protocol choices, benchmark-matched comparisons, and judge-vs-human evaluation checks.

Why This Matters For Eval Research

Protocol Takeaways

Benchmark Interpretation

  • Retrieval appears in 11.7% of hub papers (95/812); use this cohort for benchmark-matched comparisons.
  • MATH appears in 2.5% of hub papers (20/812); use this cohort for benchmark-matched comparisons.

Metric Interpretation

  • accuracy is reported in 25.9% of hub papers (210/812); compare with a secondary metric before ranking methods.
  • cost is reported in 8.7% of hub papers (71/812); compare with a secondary metric before ranking methods.

Researcher Checklist

  • Close gap on Papers with explicit human feedback. Coverage is a replication risk (13.9% vs 45% target).
  • Close gap on Papers reporting quality controls. Coverage is a replication risk (4.2% vs 30% target).
  • Tighten coverage on Papers naming benchmarks/datasets. Coverage is usable but incomplete (26.1% vs 35% target).
  • Maintain strength on Papers naming evaluation metrics. Coverage is strong (49% vs 35% target).
  • Close gap on Papers with known rater population. Coverage is a replication risk (9% vs 35% target).
  • Close gap on Papers with known annotation unit. Coverage is a replication risk (9.2% vs 35% target).

Papers with explicit human feedback

Coverage is a replication risk (13.9% vs 45% target).

Papers reporting quality controls

Coverage is a replication risk (4.2% vs 30% target).

Papers naming benchmarks/datasets

Coverage is usable but incomplete (26.1% vs 35% target).

Papers naming evaluation metrics

Coverage is strong (49% vs 35% target).

Papers with known rater population

Coverage is a replication risk (9% vs 35% target).

Papers with known annotation unit

Coverage is a replication risk (9.2% vs 35% target).

Suggested Reading Order

  1. 1. Scale Can't Overcome Pragmatics: The Impact of Reporting Bias on Vision-Language Reasoning

    High citation traction makes this a useful baseline for method and protocol context.

  2. 2. LLM Novice Uplift on Dual-Use, In Silico Biology Tasks

    High citation traction makes this a useful baseline for method and protocol context.

  3. 3. A Mixture-of-Experts Model for Multimodal Emotion Recognition in Conversations

    High citation traction makes this a useful baseline for method and protocol context.

  4. 4. SPARTA: Scalable and Principled Benchmark of Tree-Structured Multi-hop QA over Text and Tables

    Adds automatic metrics for broader coverage within this hub.

  5. 5. Discourse-Aware Dual-Track Streaming Response for Low-Latency Spoken Dialogue Systems

    Adds automatic metrics for broader coverage within this hub.

  6. 6. Why Diffusion Language Models Struggle with Truly Parallel (Non-Autoregressive) Decoding?

    Adds automatic metrics for broader coverage within this hub.

Known Limitations

  • Only 4.2% of papers report quality controls; prioritize calibration/adjudication evidence.
  • Rater population is under-specified (9% coverage).
  • Narrative synthesis is grounded in metadata and abstracts only; full-paper implementation details are not parsed.

Research Utility Links

human_eval vs llm_as_judge

both=0, left_only=9, right_only=2

0 papers use both Human Eval and Llm As Judge.

human_eval vs automatic_metrics

both=9, left_only=0, right_only=803

9 papers use both Human Eval and Automatic Metrics.

llm_as_judge vs automatic_metrics

both=2, left_only=0, right_only=810

2 papers use both Llm As Judge and Automatic Metrics.

Benchmark Brief

Retrieval

Coverage: 95 papers (11.7%)

95 papers (11.7%) mention Retrieval.

Examples: MTRAG-UN: A Benchmark for Open Challenges in Multi-Turn RAG Conversations , MoDora: Tree-Based Semi-Structured Document Analysis System , TCM-DiffRAG: Personalized Syndrome Differentiation Reasoning Method for Traditional Chinese Medicine based on Knowledge Graph and Chain of Thought

Benchmark Brief

MATH

Coverage: 20 papers (2.5%)

20 papers (2.5%) mention MATH.

Examples: Improving Parametric Knowledge Access in Reasoning Language Models , GATES: Self-Distillation under Privileged Context with Consensus Gating , Gradient Regularization Prevents Reward Hacking in Reinforcement Learning from Human Feedback and Verifiable Rewards

Benchmark Brief

MMLU

Coverage: 15 papers (1.8%)

15 papers (1.8%) mention MMLU.

Examples: Confidence-Driven Multi-Scale Model Selection for Cost-Efficient Inference , D-COT: Disciplined Chain-of-Thought Learning for Efficient Reasoning in Small Language Models , Black-Box Reliability Certification for AI Agents via Self-Consistency Sampling and Conformal Calibration

Metric Brief

accuracy

Coverage: 210 papers (25.9%)

210 papers (25.9%) mention accuracy.

Examples: InnerQ: Hardware-aware Tuning-free Quantization of KV Cache for Large Language Models , Assessing Deanonymization Risks with Stylometry-Assisted LLM Agent , Quantity Convergence, Quality Divergence: Disentangling Fluency and Accuracy in L2 Mandarin Prosody

Metric Brief

cost

Coverage: 71 papers (8.7%)

71 papers (8.7%) mention cost.

Examples: Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching , Replacing Multi-Step Assembly of Data Preparation Pipelines with One-Step LLM Pipeline Generation for Table QA , Reinforcing Real-world Service Agents: Balancing Utility and Cost in Task-oriented Dialogue

Metric Brief

latency

Coverage: 37 papers (4.6%)

37 papers (4.6%) mention latency.

Examples: Discourse-Aware Dual-Track Streaming Response for Low-Latency Spoken Dialogue Systems , Why Diffusion Language Models Struggle with Truly Parallel (Non-Autoregressive) Decoding? , InnerQ: Hardware-aware Tuning-free Quantization of KV Cache for Large Language Models

Most Cited In This Hub

Fast path to methods with the strongest citation traction in this scope.

Papers: Scale Can't Overcome Pragmatics: The Impact of Reporting Bias on Vision-Language Reasoning , LLM Novice Uplift on Dual-Use, In Silico Biology Tasks , A Mixture-of-Experts Model for Multimodal Emotion Recognition in Conversations

Most Recent

Fast path to latest protocol changes and newly published evaluation setups.

Papers: Scale Can't Overcome Pragmatics: The Impact of Reporting Bias on Vision-Language Reasoning , LLM Novice Uplift on Dual-Use, In Silico Biology Tasks , A Mixture-of-Experts Model for Multimodal Emotion Recognition in Conversations

Best Protocol Detail

Papers with explicit rater/unit metadata and quality-control signals for reproducibility.

Papers: An artificial intelligence framework for end-to-end rare disease phenotyping from clinical notes using large language models , Cross-Cultural Expert-Level Art Critique Evaluation with Vision-Language Models

Top Papers

Related Hubs