HFEPX · Eval paper review
Jaeyong Ko, Pilsung Kang, Yukyung Lee
Published
Jun 24, 2026
Citations
0
Trust level
High
Usefulness score
65/100 (Medium)
Extraction confidence
80% (High)
Derived from extracted protocol signals and abstract evidence.
Rater population
Not reported
Signals refreshed
Jun 25, 2026
This paper has useful evaluation signal, but protocol completeness is partial; pair it with related papers before deciding implementation strategy.
Use this as a practical starting point for protocol research, then validate against the original paper.
Best use
Secondary protocol comparison source
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.
Useful as a secondary reference; validate protocol details against neighboring papers.
If you are doing eval pipeline work, start here
Large language models (LLMs) reach high accuracy in mathematical reasoning, but individual traces on the same problem diverge; some arrive at the correct answer while others fail. Prior work analyzes failure at the step, chunk, or sentence level, or at tokens where failure has already occurred. Neither identifies the precise token that triggers the shift toward failure. We introduce the cliff token, a token where the token-wise potential drops significantly under an adaptive threshold that scales with the local token-wise potential, based on a one-sided two-proportion z-test. Across seven models and three mathematical reasoning benchmarks (GSM1K, MATH500, AIME 2025), cliff tokens act as failure triggers; deleting the first cliff token and resampling recovers pass@64 to 1.0, while keeping it limits recovery to between 0.71 and 1.00. We further introduce a cliff taxonomy of deterministic, uncertain, and sampled-off cliffs, defined by greedy choice and token entropy. Each type has distinct probabilistic characteristics, and the taxonomy generalizes across model scales. Finally, we validate the taxonomy via single-token preference optimization at cliff positions (Cliff-DPO). Trained on GSM8K, Cliff-DPO improves accuracy across benchmarks by up to +6.6. Optimizing at uncertain and sampled-off cliffs improves reasoning, while deterministic cliffs do not.
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.
Pairwise Preference
Directly usable for protocol triage.
"Large language models (LLMs) reach high accuracy in mathematical reasoning, but individual traces on the same problem diverge; some arrive at the correct answer while others fail."
Automatic Metrics
Includes extracted eval setup.
"Large language models (LLMs) reach high accuracy in mathematical reasoning, but individual traces on the same problem diverge; some arrive at the correct answer while others fail."
Not reported
No explicit QC controls found.
"Large language models (LLMs) reach high accuracy in mathematical reasoning, but individual traces on the same problem diverge; some arrive at the correct answer while others fail."
MATH 500, GSM8K, AIME
Useful for quick benchmark comparison.
"Across seven models and three mathematical reasoning benchmarks (GSM1K, MATH500, AIME 2025), cliff tokens act as failure triggers; deleting the first cliff token and resampling recovers pass@64 to 1.0, while keeping it limits recovery to between 0.71 and 1.00."
Accuracy, Pass@64
Useful for evaluation criteria comparison.
"Large language models (LLMs) reach high accuracy in mathematical reasoning, but individual traces on the same problem diverge; some arrive at the correct answer while others fail."
Large language models (LLMs) reach high accuracy in mathematical reasoning, but individual traces on the same problem diverge; some arrive at the correct answer while others fail.
Based on abstract + metadata only. Check the source paper before making high-confidence protocol decisions.
Human feedback protocol is explicit
Detected: Pairwise Preference
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: MATH-500, GSM8K, AIME
Metric reporting is present
Detected: accuracy, pass@64
