Dataset Viewer
Auto-converted to Parquet Duplicate
rule_id
stringlengths
14
43
label
stringlengths
9
65
family
stringclasses
5 values
code
stringlengths
21
163
train_rank_odd
odd ranks
static_rank
return card.rank % 2 == 1
train_suit_hearts
hearts only
static_suit_color
return card.suit == "hearts"
train_combo_black_low
black ranks ace through 7
static_combo
return card.color == "black" and card.rank <= 7
train_seq_same_color
same color as previous card
previous_card
if not mainline: return True return card.color == mainline[-1].color
train_pair_colors
colors appear in pairs and each pair changes color
pair_position
if not mainline: return True if len(mainline) % 2 == 0: return card.color != mainline[-1].color return card.color == mainline[-1].color
train_rank_low_1_6
ranks ace through 6
static_rank
return card.rank <= 6
train_suit_no_diamonds
anything except diamonds
static_suit_color
return card.suit != "diamonds"
train_combo_red_face
red face cards
static_combo
return card.color == "red" and card.rank >= 11
train_pair_parity
rank parity appears in pairs and each pair changes parity
pair_position
if not mainline: return True if len(mainline) % 2 == 0: return (card.rank % 2) != (mainline[-1].rank % 2) return (card.rank % 2) == (mainline[-1].rank % 2)
train_rank_high_8_13
ranks 8 through king
static_rank
return card.rank >= 8
train_suit_black
black cards
static_suit_color
return card.color == "black"
train_combo_red_odd_black_even
red odd cards and black even cards
static_combo
if card.color == "red": return card.rank % 2 == 1 return card.rank % 2 == 0
train_seq_same_suit
same suit as previous card
previous_card
if not mainline: return True return card.suit == mainline[-1].suit
train_rank_middle_4_10
ranks 4 through 10
static_rank
return 4 <= card.rank <= 10
train_suit_hearts_or_clubs
hearts or clubs
static_suit_color
return card.suit in {"hearts", "clubs"}
train_rank_non_prime
non-prime ranks
static_rank
primes = {2, 3, 5, 7, 11, 13} return card.rank not in primes
train_seq_same_rank
same rank as previous card
previous_card
if not mainline: return True return card.rank == mainline[-1].rank
train_rank_kings
only kings
static_rank
return card.rank == 13
train_seq_same_parity
same rank parity as previous card
previous_card
if not mainline: return True return (card.rank % 2) == (mainline[-1].rank % 2)
train_seq_strict_increase
rank strictly increases
previous_card
if not mainline: return True return card.rank > mainline[-1].rank
train_seq_strict_decrease
rank strictly decreases
previous_card
if not mainline: return True return card.rank < mainline[-1].rank
train_seq_non_increasing
rank does not increase
previous_card
if not mainline: return True return card.rank <= mainline[-1].rank
train_seq_rank_diff_exactly_one
rank changes by exactly 1
previous_card
if not mainline: return True return abs(card.rank - mainline[-1].rank) == 1
train_seq_rank_diff_at_most_one
rank changes by at most 1
previous_card
if not mainline: return True return abs(card.rank - mainline[-1].rank) <= 1
train_seq_rank_diff_at_most_two
rank changes by at most 2
previous_card
if not mainline: return True return abs(card.rank - mainline[-1].rank) <= 2
train_seq_rank_diff_at_least_three
rank changes by at least 3
previous_card
if not mainline: return True return abs(card.rank - mainline[-1].rank) >= 3
train_seq_same_low_high_group
same low/high rank group as previous card
previous_card
if not mainline: return True return (card.rank >= 8) == (mainline[-1].rank >= 8)
train_seq_different_low_high_group
different low/high rank group than previous card
previous_card
if not mainline: return True return (card.rank >= 8) != (mainline[-1].rank >= 8)
train_seq_same_face_status
same face/number status as previous card
previous_card
if not mainline: return True return (card.rank >= 11) == (mainline[-1].rank >= 11)
train_seq_red_down_black_up
red previous goes down, black previous goes up
previous_card
if not mainline: return 5 <= card.rank <= 9 prev = mainline[-1] if prev.color == "red": return card.rank <= prev.rank return card.rank >= prev.rank
train_seq_suit_cycle_dchs
suits cycle diamonds, clubs, hearts, spades
previous_card
if not mainline: return True order = ["diamonds", "clubs", "hearts", "spades"] idx = order.index(mainline[-1].suit) return card.suit == order[(idx + 1) % 4]
train_seq_face_diff_color_number_same_color
faces require different color, numbers require same color
previous_card
if not mainline: return True prev = mainline[-1] if prev.rank >= 11: return card.color != prev.color return card.color == prev.color
train_seq_share_suit_or_parity
same suit or same rank parity as previous card
previous_card
if not mainline: return True prev = mainline[-1] return card.suit == prev.suit or (card.rank % 2) == (prev.rank % 2)
train_suit_clubs
clubs only
static_suit_color
return card.suit == "clubs"
train_suit_diamonds
diamonds only
static_suit_color
return card.suit == "diamonds"
train_suit_no_hearts
no hearts
static_suit_color
return card.suit != "hearts"
train_suit_hearts_or_spades
hearts or spades
static_suit_color
return card.suit in {"hearts", "spades"}
train_rank_even_low_set
even ranks up to eight
static_rank
return card.rank in {2, 4, 6, 8}
train_rank_low_1_3
ranks one to three
static_rank
return card.rank <= 3
train_rank_high_10_13
ranks ten and up
static_rank
return card.rank >= 10
train_rank_band_7_11
ranks seven to eleven
static_rank
return 7 <= card.rank <= 11
train_combo_black_high
black high cards
static_combo
return card.color == "black" and card.rank >= 10
train_seq_different_mid_group
alternate low and mid-high vs previous (threshold five)
previous_card
if not mainline: return True return (card.rank >= 5) != (mainline[-1].rank >= 5)
train_pair_suits
pairs share suit (opposite phase)
pair_position
if not mainline: return True if len(mainline) % 2 == 0: return card.suit == mainline[-1].suit return card.suit != mainline[-1].suit
train_pair_ranks
pairs share rank (opposite phase)
pair_position
if not mainline: return True if len(mainline) % 2 == 0: return card.rank == mainline[-1].rank return card.rank != mainline[-1].rank
train_rank_even_red
red even ranks
static_combo
return card.color == "red" and card.rank % 2 == 0
train_rank_even_black
black even ranks
static_combo
return card.color == "black" and card.rank % 2 == 0
train_rank_even_or_king
even ranks or kings
static_rank
return card.rank % 2 == 0 or card.rank == 13
train_rank_odd_primes
odd primes
static_rank
return card.rank in {3, 5, 7, 11, 13}
train_seq_different_high_group_10
alternate low and high vs previous (threshold ten)
previous_card
if not mainline: return True return (card.rank >= 10) != (mainline[-1].rank >= 10)
train_seq_group_hc_ds
alternate between hearts+clubs and diamonds+spades
previous_card
if not mainline: return True group_a = {"hearts", "clubs"} return (card.suit in group_a) != (mainline[-1].suit in group_a)
train_seq_face_same_color_number_diff_color
after a face card same color, after a number card different color
previous_card
if not mainline: return True prev = mainline[-1] if prev.is_face: return card.color == prev.color return card.color != prev.color

eleusis-hf-rules

Rule library for the eleusis inductive-reasoning environment. Each rule is a Python predicate over (card, mainline).

  • test (26 rules): the rules of the HF eleusis benchmark, verified behaviorally identical to the benchmark repository's library.
  • train (52 rules): a training set matched to the test split — exactly 2x its family mix and closely matched acceptance-rate distribution — with zero behavioral overlap with the test rules (verified by extension-signature comparison over a probe battery).

Columns: rule_id, label, family, code.

Downloads last month
194