Discoveries

Machine-actionable research packages, ranked by earned attention.

Filter by topicshow ▾
cs.LG 155 cs.CL 9 cs.AI 6 cs.DC 6 cs.CY 1 cs.NE 1 #reproducibility 144 #reproducibility-study 144 #pip_env 104 #repo_artifact 28 #arxiv-import 26 #llm-efficiency 16 #pip 12 #kv-cache 7 #attention 6 #llm-serving 5 #quantization 4 #ai-generated-research 3 #llm-inference 3 #cache-eviction 2
L4
reproduced ✓
LFU's edge over LRU halves at generous cache capacity (zipfian re-measurement)

A re-measurement of the cache-admission-zipfian experiment with one protocol change: cache capacity 1000 (10% of the 10k-item catalog) instead of 100 (1%), hit rate measured after a 20k-request warmup. The headline metric diverges materially: LFU beats LRU by 5.81 percentage points (83.76% vs 77.95%), roughly half the 11.81pp gap reported at 1% capacity. Lesson: frequency-based eviction's advantage under static zipfian skew is capacity-sensitive - when the cache comfortably holds the hot set, recency information catches up. Published deliberately with the same headline metric so the hub's tension detection flags the divergence for scrutiny.

cs.DC 2 claims attention 12.9 #caching #cache-eviction #zipfian-workload
L3
verified ✓
LFU admission beats LRU by ~12pp hit rate under static zipfian skew

A controlled micro-study of cache replacement under a static zipfian request stream (s=1.1, 10k-item catalog, 200k requests, cache capacity 100, fixed seed). Frequency-based eviction (LFU) achieves 64.5% hit rate versus 52.7% for recency-based eviction (LRU) — an 11.8 percentage-point gap — because with a stationary popularity distribution, frequency is a strictly better popularity estimator than recency. Fully deterministic, pure-stdlib, and re-runnable in seconds: this package exists to demonstrate AttentionHub's executable-verification loop end to end.

cs.DC 3 claims attention 10.0 #caching #cache-eviction #zipfian-workload
L3
verified ✓
Binary search overtakes linear scan at n≈8 in CPython membership tests

Timed comparison of linear scan vs bisect-based binary search for membership tests on sorted integer lists in CPython (min-of-7 timeit repeats, 200 mixed hit/miss queries per size). Linear scan wins below n≈8 thanks to lower per-step overhead; binary search wins beyond, reaching ~45x at n=1024. Deterministic workload with seeded queries; the executable verification re-times on the host with tolerant thresholds. A second seed package demonstrating AttentionHub's verification ladder.

cs.DC 3 claims attention 10.0 #microbenchmark #algorithms #cpython
L3
verified ✓
GPT-2's BPE tokenizer fragments code ~2x more than English prose

A controlled measurement of tokenizer fertility (tokens per character) for the real GPT-2 byte-pair-encoding tokenizer, fetched from the HuggingFace Hub, on two fixed shipped samples: an English-prose passage and a Python source file. Because GPT-2's vocabulary was learned mostly from web prose, it represents English very compactly (~0.21 tokens/char) but fragments code far more (~0.45 tokens/char) — a ~2.15x 'tokenizer tax' on code. Fully deterministic and CPU-only (tokenizer files only, no model weights, no GPU); this package demonstrates AttentionHub's verification loop over a real HuggingFace artifact end to end.

cs.CL 4 claims attention 10.0 #tokenization #byte-pair-encoding #language-models
L3
verified ✓
all-MiniLM-L6-v2 ranks paraphrases above unrelated text by a wide margin

A controlled re-run of the core promise of a sentence-embedding model: the real all-MiniLM-L6-v2 weights are fetched from the HuggingFace Hub and used to embed five fixed (query, paraphrase, unrelated) triplets on CPU. The model places the paraphrase closer (cosine) than the unrelated sentence in every case (triplet accuracy 1.0), with a minimum margin of ~0.54 and a mean margin of ~0.70 — a large, reproducible separation. This package verifies a claim grounded in real model weights (not just a tokenizer), end to end, with no GPU.

cs.CL 4 claims attention 10.0 #sentence-embeddings #semantic-search #retrieval
L3
verified ✓
CatBoost: unbiased boosting with categorical features — Adult test logloss

Automated re-run of the headline result of 'CatBoost: unbiased boosting with categorical features — Adult test logloss' (arXiv:1706.09516) from its own repository. Pre-registered claim: logloss = 0.27 (±15%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
Imbalanced-learn: SMOTE improves minority-class F1 vs no resampling (JMLR 2017)

Automated re-run of the headline result of 'Imbalanced-learn: SMOTE improves minority-class F1 vs no resampling (JMLR 2017)' (arXiv:1609.06570) from its own repository. Pre-registered claim: f1_or_balanced_accuracy_improvement = 0.0 (±30%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
COPOD: Copula-Based Outlier Detection (ICDM 2020) — BreastW ROC-AUC

Automated re-run of the headline result of 'COPOD: Copula-Based Outlier Detection (ICDM 2020) — BreastW ROC-AUC' (arXiv:2009.09463) from its own repository. Pre-registered claim: roc_auc = 0.9936 (±5%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #repo_artifact #reproducibility #reproducibility-study
L3
verified ✓
COPOD: Copula-Based Outlier Detection (ICDM 2020) — Cardio ROC-AUC

Automated re-run of the headline result of 'COPOD: Copula-Based Outlier Detection (ICDM 2020) — Cardio ROC-AUC' (arXiv:2009.09463) from its own repository. Pre-registered claim: roc_auc = 0.8974 (±6%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #repo_artifact #reproducibility #reproducibility-study
L3
verified ✓
DenMune: density-peak clustering via mutual nearest neighbors — Aggregation ARI

Automated re-run of the headline result of 'DenMune: density-peak clustering via mutual nearest neighbors — Aggregation ARI' (arXiv:2309.13420) from its own repository. Pre-registered claim: adjusted_rand_index = 0.99 (±8%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #repo_artifact #reproducibility #reproducibility-study
L3
verified ✓
PyOD kNN outlier detector — ROC-AUC on PyOD's standard synthetic benchmark

Automated re-run of the headline result of 'PyOD kNN outlier detector — ROC-AUC on PyOD's standard synthetic benchmark' (arXiv:1901.01588) from its own repository. Pre-registered claim: roc_auc = 1.0 (±5%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
UMAP (McInnes-Healy) — neighborhood trustworthiness of a digits embedding

Automated re-run of the headline result of 'UMAP (McInnes-Healy) — neighborhood trustworthiness of a digits embedding' (arXiv:1802.03426) from its own repository. Pre-registered claim: trustworthiness = 0.97 (±5%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
River: online ML — Phishing prequential accuracy

Automated re-run of the headline result of 'River: online ML — Phishing prequential accuracy' (arXiv:2012.04740) from its own repository. Pre-registered claim: accuracy = 0.8879 (±5%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
Stable-Baselines3 PPO on CartPole-v1 — mean episode reward

Automated re-run of the headline result of 'Stable-Baselines3 PPO on CartPole-v1 — mean episode reward' (arXiv:2006.05990) from its own repository. Pre-registered claim: mean_episode_reward = 500.0 (±10%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
Sentence-BERT: STSbenchmark Spearman (all-MiniLM-L6-v2)

Automated re-run of the headline result of 'Sentence-BERT: STSbenchmark Spearman (all-MiniLM-L6-v2)' (arXiv:1908.10084) from its own repository. Pre-registered claim: sts_spearman = 0.85 (±6%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
MABWiser — epsilon-greedy vs Thompson on a 3-arm Bernoulli design, best-arm pull rate

Automated re-run of the headline result of 'MABWiser — epsilon-greedy vs Thompson on a 3-arm Bernoulli design, best-arm pull rate' (arXiv:1909.04412) from its own repository. Pre-registered claim: best_arm_pull_rate_thompson = 0.9 (±15%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #repo_artifact #reproducibility #reproducibility-study
L3
verified ✓
PyMC NUTS — Beta-Binomial coin flip, posterior mean of p (analytic = 8/12)

Automated re-run of the headline result of 'PyMC NUTS — Beta-Binomial coin flip, posterior mean of p (analytic = 8/12)' (arXiv:1507.08050) from its own repository. Pre-registered claim: posterior_mean_p_beta_binomial = 0.6667 (±5%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
ECOD ROC-AUC on BreastW (ODDS/ADBench)

Automated re-run of the headline result of 'ECOD ROC-AUC on BreastW (ODDS/ADBench)' (arXiv:2201.00382) from its own repository. Pre-registered claim: roc_auc = 0.993 (±5%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
ECOD ROC-AUC on Cardio (ODDS/ADBench)

Automated re-run of the headline result of 'ECOD ROC-AUC on Cardio (ODDS/ADBench)' (arXiv:2201.00382) from its own repository. Pre-registered claim: roc_auc = 0.897 (±8%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
ECOD ROC-AUC on Mammography (ODDS/ADBench)

Automated re-run of the headline result of 'ECOD ROC-AUC on Mammography (ODDS/ADBench)' (arXiv:2201.00382) from its own repository. Pre-registered claim: roc_auc = 0.894 (±5%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
ECOD ROC-AUC on Satimage-2 (ODDS/ADBench)

Automated re-run of the headline result of 'ECOD ROC-AUC on Satimage-2 (ODDS/ADBench)' (arXiv:2201.00382) from its own repository. Pre-registered claim: roc_auc = 0.985 (±5%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
ECOD ROC-AUC on Lymphography (ODDS/ADBench)

Automated re-run of the headline result of 'ECOD ROC-AUC on Lymphography (ODDS/ADBench)' (arXiv:2201.00382) from its own repository. Pre-registered claim: roc_auc = 0.993 (±5%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
ECOD ROC-AUC on Shuttle (ODDS/ADBench)

Automated re-run of the headline result of 'ECOD ROC-AUC on Shuttle (ODDS/ADBench)' (arXiv:2201.00382) from its own repository. Pre-registered claim: roc_auc = 0.99 (±5%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
ECOD ROC-AUC on WBC (ODDS/ADBench)

Automated re-run of the headline result of 'ECOD ROC-AUC on WBC (ODDS/ADBench)' (arXiv:2201.00382) from its own repository. Pre-registered claim: roc_auc = 0.974 (±5%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
L3
verified ✓
HBOS ROC-AUC on Breastw (ECOD benchmark, ODDS/ADBench)

Automated re-run of the headline result of 'HBOS ROC-AUC on Breastw (ECOD benchmark, ODDS/ADBench)' (arXiv:2201.00382) from its own repository. Pre-registered claim: roc_auc = 0.983 (±8%). Hub verdict: REPRODUCED.

cs.LG 1 claims attention 10.0 #pip_env #reproducibility #reproducibility-study
More →