EasyDeepLearn
Reinforcement Learning · section 8 of 12

Offline RL

22 interview questions on offline rl, each answered in full. Free to read, no account needed.

What is off-policy evaluation and why is it hard?

hard
  • Off-policy evaluation estimates the value of a target policy using data collected by a different behavior policy — without deploying the new policy.
  • It's crucial in medicine, finance, and any high-stakes system where trying the new policy is risky.
  • It is hard because of distribution shift; techniques include importance sampling, doubly robust estimators, and model-based methods.
  • Variance and bias tradeoffs are severe.
#evaluation#safetyPermalink & quiz →

What is offline RL and why does it matter?

medium
  • Learn policy from fixed dataset of (s, a, r, s') collected by some behavior policy — NO env interaction during training.
  • Motivation: (1) real environments too expensive / risky to explore (healthcare, autonomous, industrial).
  • (2) Leverage huge existing logs (recommendation clicks, robot demos, dialog logs).
  • (3) Safer than online.
  • Foundation of practical industrial RL.
  • D4RL is standard benchmark suite.

Why is distribution shift the core challenge in offline RL?

hard
  • Policy π_θ may query Q at (s, a) never seen in data → Q's extrapolation is nonsense → policy exploits Q's overestimation of OOD actions → learned policy fails.
  • Naive off-policy methods (SAC on offline data) diverge catastrophically.
  • Fixes: constrain policy near behavior (BCQ), pessimistic Q (CQL), model-based with uncertainty (MOReL, COMBO), decision-transformer (no Q at all).

BCQ — Batch-Constrained Q-Learning.

hard
  • Fujimoto et al. 2019.
  • Constrain policy to only take actions similar to those in the data.
  • Uses a conditional VAE trained on (s, a) pairs to sample actions π(as){\pi}(a \mid s) close to behavior.
  • Q-learning then only over these sampled 'safe' actions.
  • Effectively KL-constrained policy improvement without explicit KL.
  • Modern successors: TD3+BC (adds BC term), IQL (uses expectiles).

CQL — Conservative Q-Learning.

hard
  • Kumar et al. 2020.
  • Add penalty to standard Q-learning loss: expectation of Q over the policy (pushed down) minus expectation over data (pushed up) → pessimistic on OOD actions.
  • Prevents Q from over-estimating outside data.
  • Widely used in modern offline RL benchmarks.
  • Result: policy converges to a lower-bound on true Q → conservative, safe behavior.
  • Simple to implement on top of SAC.

IQL — Implicit Q-Learning.

hard
  • Kostrikov et al. 2022.
  • Avoids explicit policy constraint or Q pessimism.
  • Uses expectile regression to approximate maxa\operatorname{max}_{a} Q(s, a) without evaluating Q at unseen actions — only uses actions from the data.
  • Extract policy via advantage-weighted regression on behavior policy.
  • Simpler, more stable than BCQ / CQL.
  • Often the modern default for offline RL benchmarks in 2023+.

TD3+BC — the simplest offline RL baseline.

medium
  • Fujimoto & Gu 2021.
  • TD3 (actor-critic) + BC (behavior cloning) regularizer added to policy loss: min E[-Q(s, π(s))] + λ (π(s)    adata)2({\pi}(s)\; - \;a_{\mathrm{data}})^{2}.
  • Encourages policy to stay close to data actions.
  • Extremely simple (add one term), strong performance on D4RL.
  • Foundation demonstrating that a small BC regularization is often enough for offline RL — sophisticated methods (CQL, IQL) add marginal gains.

Behavior Cloning (BC) — when does it work?

medium
  • Supervised learning: fit π(as){\pi}(a \mid s) to (s, a) pairs from expert demos via MLE.
  • Works when: (1) expert data is high quality, (2) test distribution matches expert distribution (no drift), (3) enough coverage.
  • Fails: (1) compounding errors under distribution shift (agent drifts, sees unseen states, makes worse mistakes).
  • Solutions: DAgger (query expert in new states), residual RL, BC + RL fine-tuning.
#offline-rl#imitationPermalink & quiz →

DAgger — Dataset Aggregation for imitation learning.

medium
  • Ross et al. 2011.
  • Iterative BC: (1) train π on expert data, (2) roll out π in env, (3) query expert for correct action at every state visited, (4) add (state, expert-action) to dataset, (5) retrain.
  • Fixes BC's compounding errors by getting expert labels on the agent's own state distribution.
  • Requires access to expert during training.
  • Foundation of interactive imitation learning.
#offline-rl#imitationPermalink & quiz →

Inverse Reinforcement Learning (IRL) — the setup.

hard
  • Given expert demonstrations, infer the reward function that makes the expert optimal.
  • Ill-posed (many rewards consistent).
  • Approaches: (1) Feature-matching: match expected features under learner and expert.
  • (2) Max-entropy IRL (Ziebart 2008): among consistent rewards, prefer max-entropy.
  • (3) GAIL (Ho & Ermon 2016): GAN-like adversarial training — discriminator judges expert vs learner.
  • Foundation of modern imitation learning + robotics.
#offline-rl#imitationPermalink & quiz →

GAIL — Generative Adversarial Imitation Learning.

hard
  • Ho & Ermon 2016.
  • Similar to GAN: generator = policy π, discriminator D distinguishes agent trajectories from expert trajectories.
  • Train π to fool D via policy gradient (log D as reward).
  • Recovers expert behavior without explicit reward.
  • Extends to continuous control + high-dim states.
  • Foundation of adversarial imitation learning.
  • Modern extensions: SQIL, ValueDICE, AdVIL.
#offline-rl#imitationPermalink & quiz →

Reward hacking in RLHF — canonical examples.

hard
  • Model finds ways to score high on r_φ without being genuinely helpful: (1) sycophancy: agree with user regardless of truth.
  • (2) Length bias: longer responses often scored higher by human raters, so model over-verbose.
  • (3) Excessive hedging (safety over-refusal).
  • (4) Format exploitation (bullets look thorough).
  • (5) Adversarial phrases that game the model.
  • Fixes: iterative reward model updates, length-normalized rewards, adversarial red-team, RLAIF (AI feedback).
#llm#alignment#safetyPermalink & quiz →

Constitutional AI — mechanism.

hard
  • Anthropic 2022.
  • Two phases: (1) SL from AI feedback: model critiques + revises its own responses per written constitution (helpful, harmless principles); train on (prompt, revised response).
  • (2) RL from AI feedback: LLM judge scores pairs of responses on constitutional principles; train reward model + PPO.
  • Advantage: avoids exposing human labelers to harmful content.
  • Foundation of Claude's training.
  • Modern LLM alignment default.
#llm#alignment#safetyPermalink & quiz →

How does the KL penalty interact with reward hacking?

hard
  • KL to πref{\pi}_{\mathrm{ref}} bounds how far model can drift → limits reward hacking of imperfect r_φ.
  • But: (1) large model can find hacks within KL budget (spurious  features  that  both  rφ  likes  AND  πref  probably  would  generate)(\mathrm{spurious}\;\mathrm{features}\;\mathrm{that}\;\mathrm{both}\;r{\varphi}\;\mathrm{likes}\;\mathrm{AND}\;{\pi}_{\mathrm{ref}}\;\mathrm{probably}\;\mathrm{would}\;\mathrm{generate}).
  • (2) β must be tuned per task — too high stalls learning.
  • (3) Better: iterative RM updates (relabel new-policy responses), adversarial reward shaping.
  • (4) Modern practice: KL as safety floor, not primary defense against hacking.
#llm#alignment#safetyPermalink & quiz →

Safe RL — how do you constrain during learning?

hard
  • (1) CMDP (Constrained MDP): maximize reward subject to E[Σ  ct]E[{\Sigma}\;c_{t}] ≤ threshold.
  • Solved via Lagrangian or projected gradient.
  • (2) Safe exploration: pretrain with safe demonstrations, use safety layers.
  • (3) Reachability analysis.
  • (4) Reward shaping with constraint violations.
  • (5) Test-time verification: reject actions predicted to violate constraints.
  • Standard in robotics, autonomous vehicles, medical RL — high-stakes safety-critical domains.

Sycophancy in RLHF — what causes it and how to fix?

hard
  • Model tells users what they want to hear rather than truth.
  • Cause: human labelers prefer agreeable responses; RM learns to reward agreement patterns.
  • Fixes: (1) explicitly label sycophancy in preference data.
  • (2) Train with adversarial 'test' prompts asking for wrong assertions.
  • (3) Include truthfulness in constitutional principles.
  • (4) Ensemble RMs to detect disagreement.
  • Ongoing challenge in modern LLM alignment.
#llm#alignment#safetyPermalink & quiz →

How does RL contribute to jailbreak defense?

hard
  • (1) Safety-tuned RM penalizes harmful outputs.
  • (2) Adversarial RLHF: red-team probes for jailbreaks, add examples to preference data.
  • (3) Refusal tuning: SFT + RLHF on 'appropriate refusal' patterns.
  • (4) Multi-round RLAIF on constitutional harmful principles.
  • Trade-off: too aggressive → over-refusal; too lax → harmful outputs.
  • Modern practice: continuous red-teaming with automated attacks (GCG) as part of RM training loop.
#llm#alignment#safetyPermalink & quiz →

How is o1-style reasoning safer or riskier?

hard
  • Safer: (1) longer reasoning gives more opportunities to catch errors + reject harmful requests.
  • (2) Explicit deliberation can invoke safety principles.
  • Riskier: (3) can 'think its way' around safety training if not RL-trained on refusal.
  • (4) Chain-of-thought exposes reasoning that could be jailbroken.
  • (5) Increased test-time compute → some safety mitigations at inference.
  • Modern practice: RL train reasoning models with explicit safety verifier reward.
#llm#alignment#safetyPermalink & quiz →

Why is RL hard for autonomous driving?

hard
  • (1) Safety: cannot explore in real traffic.
  • (2) Long-tail rare events matter (fatal edge cases).
  • (3) Multi-agent interactions (other drivers, pedestrians) with non-stationary policies.
  • (4) Partial observability.
  • (5) Reward specification difficulty.
  • Most industry uses SL / imitation learning + planning stack (Waymo, Cruise).
  • RL used mostly in simulation for policy fine-tuning, closed-loop training.
  • Recent: Tesla's neural planners are hybrid IL + RL.
#applications#safetyPermalink & quiz →

RL in healthcare — key challenges.

hard
  • (1) Sample efficiency (real trials cost billions).
  • (2) Safety absolute — can't try random treatments.
  • (3) Off-policy evaluation on historical patient data (dosing decisions from EHR) is standard.
  • (4) Confounders in observational data — need causal RL.
  • (5) Reward specification: what is 'good outcome' over years?
  • Applications: dosing optimization (mechanical ventilation, sepsis), personalized treatment.
  • Deployed in advisory tools only; not autonomous treatment selection.
#applications#safetyPermalink & quiz →

Interview: 'how do you make an RL policy safe to deploy?'

hard
  • (1) Simulation validation across held-out scenarios + adversarial tests.
  • (2) Formal / conservative bounds where possible (CMDP).
  • (3) Safety layer: monitor / veto extreme actions at runtime.
  • (4) Human-in-the-loop for edge cases.
  • (5) Shadow deployment first (log-only), then gradual rollout with A/B guardrails.
  • (6) Continuous monitoring: distribution shift, reward drift, action distribution, incident rate.
  • (7) Rollback plan.
  • Safety-critical RL is engineering-heavy.
#interview#safetyPermalink & quiz →

Safety verifier reward — what and how?

hard
  • Automatic reward from external verifier that judges safety: (1) content filter classifier flags harmful outputs.
  • (2) LLM judge with harm rubric.
  • (3) Rule-based checks (PII leakage, illegal content).
  • Combined with helpfulness reward → multi-objective RL.
  • Alignment tax risk if safety over-weighted.
  • Modern: 'safety as red-team' — actively probe model with attacks, RL trains against them (Anthropic's HH pipeline).
#llm#alignment#safetyPermalink & quiz →

Practise Reinforcement Learning