DeepSeek-Prover-V1.5-RL+RMaxTS是一个结合强化学习和搜索策略的自动定理证明系统。
问题:解方程 x² - 5x + 6 = 0
操作步骤:
Theorem quadratic : exists x : Z, x^2 - 5*x + 6 = 0.
deepseek-prover --problem "quadratic" --strategy rmaxts --max-depth 10
问题:证明直角三角形的斜边平方等于两直角边平方和
操作步骤:
Theorem Pythagoras : forall a b c : nat, right_triangle a b c -> a^2 + b^2 = c^2.
from deepseek import Prover
prover = Prover(geometry_mode=True)
prover.prove("Pythagoras", strategy="RL+RMaxTS")
问题:证明若p是素数,a不被p整除,则a^(p-1) ≡ 1 mod p
操作步骤:
theorem fermat_little : ∀ p : ℕ, prime p → ∀ a : ℕ, ¬p ∣ a → a^(p-1) % p = 1 := by
deepseek-prover --domain number_theory --timeout 600
问题:求f(x) = x³·sin(x)的导数
操作步骤:
from deepseek.calculus import Differentiator
diff = Differentiator()
expr = "x**3 * sin(x)"
result = diff.compute(expr, strategy="symbolic+rmaxts")
问题:计算10人排成圆形队列的不同排列数
操作步骤:
Lemma circular_perm : forall n : nat, n > 0 -> factorial(n-1) = circular_permutations(n).
deepseek-prover --domain combinatorics --strategy counting
问题:计算二项分布B(n,p)的期望值
操作步骤:
def binomial_expectation : [X] = n * p := by
prover.set_context("probability")
prover.use_lemma("linearity_of_expectation")
问题:证明实数轴R在标准拓扑下是连通的
操作步骤:
Theorem R_connected : connected (Space.mk ℝ (metric_topology ℝ)).
deepseek-prover --domain topology --strategy connectivity
问题:证明 (P→Q) ∧ (Q→R) ⇒ (P→R)
操作步骤:
Lemma syllogism : forall P Q R : Prop, (P -> Q) /\ (Q -> R) -> (P -> R).
prover.apply_tactic("natural_deduction")
prover.intro_hypotheses()
问题:证明rank(A+B) ≤ rank(A) + rank(B)
操作步骤:
theorem rank_sum : ∀ A B : matrix, rank(A + B) ≤ rank(A) + rank(B) := by
deepseek-prover --library linear_algebra.rank_properties
问题:解方程 y’ + P(x)y = Q(x)
操作步骤:
from deepseek.ode import FirstOrderSolver
solver = FirstOrderSolver()
solution = solver.solve("dy/dx + P(x)y = Q(x)", method="integrating_factor")
--domain
参数指定数学领域,激活对应策略库