EasyDeepLearn
Reinforcement Learning · section 10 of 12

RLHF & LLM alignment

24 interview questions on rlhf & llm alignment, each answered in full. Free to read, no account needed.

What is RLHF and how does it fit into training an LLM?

hard
  • RLHF (Reinforcement Learning from Human Feedback) fine-tunes a pretrained language model to align with human preferences.
  • Steps: (1) supervised fine-tuning on curated demonstrations; (2) train a reward model on human preference pairs; (3) optimize the LLM policy against the reward model using PPO, with a KL penalty to stay close to the reference policy.
  • Recent alternatives: DPO removes the RL step by optimizing preferences directly.
#llm#alignmentPermalink & quiz →

RLHF pipeline — the three stages in detail.

hard
  • (1) SFT: fine-tune pretrained LM on curated (prompt, ideal response) demos → gives baseline capable model.
  • (2) Reward Modeling: collect preference pairs (prompt, response A vs B, human picks winner); train reward model r_φ(x, y) via Bradley-Terry loss: -log σ(r(x,  yw)    r(x,  yl)){\sigma}(r(x, \;y_{w})\; - \;r(x, \;y_{l})).
  • (3) RL: fine-tune π_θ against r_φ with PPO + KL penalty to reference πref{\pi}_{\mathrm{ref}} (β  KL(πθ    πref))({\beta}\;\operatorname{KL}({\pi}{\theta}\; \mid \mid \;{\pi}_{\mathrm{ref}})) to prevent reward hacking + preserve capabilities.
#llm#alignmentPermalink & quiz →

Bradley-Terry model in reward modeling.

hard
  • Probability preferred response yw  >  yly_{w}\; > \;y_{l} given prompt x: P(yw  >  yl    x)  =  σ(r(x,  yw)    r(x,  yl))P(y_{w}\; > \;y_{l}\; \mid \;x)\; = \;{\sigma}(r(x, \;y_{w})\; - \;r(x, \;y_{l})).
  • Loss: -log σ(r(x,  yw)    r(x,  yl)){\sigma}(r(x, \;y_{w})\; - \;r(x, \;y_{l})) over preference dataset.
  • Reward model r_φ is scalar-output classifier (usually initialized from SFT model + linear head).
  • Modern reward models: hundreds of thousands of preference pairs; usually much more efficient than expected due to LLM capabilities.
#llm#alignmentPermalink & quiz →

Why does RLHF include a KL penalty?

hard
  • PPO fine-tunes against learned reward model r_φ, which is imperfect and prone to reward hacking (model  finds  spurious  features  that  rφ  overestimates)(\mathrm{model}\;\mathrm{finds}\;\mathrm{spurious}\;\mathrm{features}\;\mathrm{that}\;r{\varphi}\;\mathrm{overestimates}).
  • KL penalty β KL(πθ    πref)\operatorname{KL}({\pi}{\theta}\; \mid \mid \;{\pi}_{\mathrm{ref}}) keeps π_θ close to SFT reference → limits how badly model can exploit reward.
  • Also preserves capabilities from pretraining.
  • Choice of β matters: too high → no learning; too low → catastrophic drift.
  • Standard β = 0.01-0.1 for chat models.
#llm#alignmentPermalink & quiz →

DPO vs PPO for RLHF — practical comparison.

hard
  • PPO: needs reward model + rollouts + KL + value function; complex + unstable.
  • DPO: single stage on preference pairs, no RM or PPO, much simpler + more stable.
  • Empirically comparable or better than PPO for most tasks (Rafailov 2023).
  • Downsides: (1) still needs preference data.
  • (2) Off-policy in a sense (uses static preference data).
  • (3) Weaker for tasks with objective correctness (math, code) where RL from verifier is better.
  • PPO still standard at frontier labs for large models.
#llm#alignmentPermalink & quiz →

RLAIF — RL from AI Feedback.

hard
  • Replace human labelers with LLM judge for preference labels.
  • Bai et al. 2022 (Constitutional AI): iteratively use a strong LLM to critique + revise responses per a written constitution.
  • Advantages: scale (billions of preferences), consistency, cost.
  • Trade-offs: risk of amplifying LLM biases, less alignment with real user preferences.
  • Standard in modern alignment (Anthropic's Constitutional AI, Meta's Llama, Google's Gemini all use RLAIF + human).
#llm#alignmentPermalink & quiz →

RL for reasoning (o1, DeepSeek-R1) — the paradigm shift.

hard
  • Train LLM to generate long chain-of-thought reasoning by RL with automatic reward from verifier (math answer correct, code passes tests).
  • Model learns to think longer, retry, self-correct.
  • DeepSeek-R1: pure RL from base model (no SFT step) → emergent reasoning behaviors.
  • Uses GRPO (group-normalized rewards).
  • Test-time compute becomes a knob.
  • Marks shift from imitation-based to verification-based post-training.
#llm#alignmentPermalink & quiz →

How does RL train chain-of-thought reasoning?

hard
  • Reward on FINAL answer only (verifier: math correct? code passes tests?).
  • Intermediate tokens get gradient via policy gradient on the final reward, credit-assigned across the response.
  • Long CoT emerges because: (1) longer thinking gives higher chance of correct answer, (2) reward differential per token pushes toward exploration of reasoning traces.
  • Combined with KL penalty to preserve language + prevent gibberish → emergent step-by-step reasoning.
#llm#alignmentPermalink & quiz →

Verifier reward vs preference-based reward — when to use each?

medium
  • Verifier: objective correctness (math, code, structured tasks) — automatic, cheap, unambiguous, scalable.
  • Best when correctness is well-defined.
  • Preference: subjective quality (helpfulness, tone, style, creative writing) — needs human/AI labels, more ambiguous.
  • Best when 'right' is unclear.
  • Modern LLM training combines both: verifier RL for skills (o1 reasoning), preference RLHF for chat quality, safety.
#llm#alignmentPermalink & quiz →

Process reward vs outcome reward for reasoning.

hard
  • Outcome reward: only final answer scored.
  • Simple but sparse — credit assignment across long CoT is hard.
  • Process reward (PRM, Lightman et al. 2023): score each reasoning step for correctness.
  • Denser signal, better for long CoT.
  • Training PRMs requires per-step labels (expensive).
  • Modern practice: mix outcome + process; auto-generate process labels via LLM verifier or Monte Carlo (reason forward from step, check outcome distribution).
#llm#alignmentPermalink & quiz →

Iterative RLHF — why do you keep going?

hard
  • Single-round RLHF: preference data collected on SFT model → RM trained → policy trained.
  • Problem: new policy visits states RM was never trained on → RM predictions drift.
  • Fix: iterative rounds — collect new preferences on current-policy responses, update RM, redo PPO/DPO.
  • Modern frontier LLMs go through many rounds.
  • Cost: expensive labeler cycles.
  • Mitigation: mix human + RLAIF preferences.
#llm#alignmentPermalink & quiz →

What can RL do that SFT cannot?

medium
  • SFT does behavior cloning of ideal responses.
  • RL adds: (1) preference between responses (SFT can't compare).
  • (2) Learning from FAILED attempts (verifier RL, HER-style relabeling).
  • (3) Exploration of the response space beyond demos.
  • (4) Direct optimization of test-time desirable behavior (long CoT, tool use, safety).
  • (5) Scaling with test-time compute.
  • Rule: SFT gives baseline capability; RL adds targeted post-training improvements.
#llm#alignmentPermalink & quiz →

How do reward models scale with size?

hard
  • Larger RMs (closer in size to policy) give better preferences: better calibration, less easy-to-exploit gaps.
  • But: too small → doesn't understand nuance; too large → expensive per training step.
  • Modern: RM around 1/3 of policy size (7B RM for 22B policy).
  • Ensemble of RMs improves robustness against hacking.
  • Some frontier labs use LLM-as-judge (whole LLM) instead of scalar-output RM head.
#llm#alignmentPermalink & quiz →

How do you fix length bias in RLHF?

medium
  • (1) Length-normalize reward: divide by response length (log or square-root).
  • (2) Add length penalty λ · length to policy loss.
  • (3) In DPO: length-normalize log-probs.
  • (4) Curate preference pairs to be length-matched.
  • (5) Train RM to explicitly discount length (RM sees length as feature and is penalized for using it).
  • Standard fix in modern LLM training; without it, models become verbose.
#llm#alignmentPermalink & quiz →

ORPO — Odds Ratio Preference Optimization.

hard
  • Hong et al. 2024.
  • Combines SFT and preference optimization in one loss without a reference model.
  • Loss = SFT loss + λ · LORL_{\mathrm{OR}} where LORL_{\mathrm{OR}} is preference log odds ratio: -log σ(log(odds(yw)/odds(yl))\operatorname{log}(\mathrm{odds}(y_{w}) / \mathrm{odds}(y_{l}))).
  • No reference policy needed, no separate SFT stage.
  • Simpler pipeline than SFT + DPO.
  • Standard for open-source model alignment when reference model is unavailable / undesired.
#llm#alignmentPermalink & quiz →

KTO — Kahneman-Tversky Optimization.

hard
  • Ethayarajh et al. 2024.
  • Uses prospect theory (loss aversion): people care more about avoiding losses than gaining.
  • Works with unpaired (binary desirable/undesirable) labels — easier to collect than paired preferences.
  • Loss makes desirable outputs more likely + undesirable less likely, weighted asymmetrically.
  • Alternative to DPO when preference pairs are hard to get.
#llm#alignmentPermalink & quiz →

What is 'alignment tax' in RLHF?

hard
  • Fine-tuning for alignment (safety, style) degrades other capabilities.
  • E.g., RLHF for helpfulness sometimes hurts factual accuracy or reasoning.
  • Root cause: KL to reference policy is not strong enough vs conflicting reward pull; sample distribution during RL doesn't cover all training-distribution capabilities.
  • Mitigation: (1) mix pretraining tokens during RL, (2) hold-out benchmark tracking (MMLU, HumanEval), (3) stronger KL to reference, (4) mix RL with continued SFT on diverse data.
#llm#alignmentPermalink & quiz →

How does RL train tool-use in LLM agents?

hard
  • (1) Reward on task success (e.g., verifier checks correct API call result).
  • (2) Multi-step trajectories with intermediate tool calls.
  • (3) GRPO / PPO with KL penalty.
  • (4) Environments: WebArena, ALFWorld, SWE-bench.
  • (5) Format constraints: tool call syntax must be valid → additional reward shaping.
  • Modern: agentic RL is next frontier of LLM training.
  • Reward hacking risk: model finds shortcuts (e.g., fake API responses).
#llm#alignmentPermalink & quiz →

Interview: 'walk me through the tradeoffs of SFT-only vs SFT+RLHF vs SFT+DPO.'

hard
  • SFT only: fast + simple + reliable + no reward hacking.
  • Ceiling limited by demo quality; can't leverage preferences.
  • SFT+RLHF (PPO): learns from preferences + can exceed demo quality + slow + expensive + reward hacking risk + complex infra.
  • SFT+DPO: preferences without PPO complexity + faster + more stable than PPO + tie infrastructure to reference model; on-par or slightly worse than tuned PPO.
  • Modern industry mostly does SFT + DPO / IPO / KTO; frontier labs still PPO.
#interview#llm#alignmentPermalink & quiz →

ORPO vs DPO — key difference.

hard
  • DPO: needs reference policy πref{\pi}_{\mathrm{ref}} (extra memory + inference cost during training).
  • ORPO: no reference policy — uses odds ratio between chosen/rejected.
  • Simpler infra (one model instead of two).
  • Comparable performance to DPO on many benchmarks.
  • Modern choice: ORPO when reference model is inconvenient (larger models, custom architectures) or when you want simpler infra.
#llm#alignmentPermalink & quiz →

LLM agent benchmarks — what tests capability?

medium
  • (1) WebArena / VisualWebArena: browse web + execute tasks.
  • (2) ALFWorld: navigate textual household environment.
  • (3) SWE-bench: fix real GitHub issues.
  • (4) AgentBench: comprehensive tool-use.
  • (5) OSWorld: real OS interaction.
  • (6) τ-bench: customer-service dialog tasks.
  • Rule: test in realistic settings (not just multiple choice); track success rate + cost per task + safety violations.
  • Modern agent RL trained on some of these.
#llm#evaluationPermalink & quiz →

'Verifier + refiner' RL pattern for reasoning.

hard
  • (1) Sample many reasoning chains from model.
  • (2) Verifier (external tool or LLM judge) scores each.
  • (3) Best-of-N: keep highest scoring — cheap test-time scaling.
  • (4) RL fine-tune: train model to produce chains verifier likes.
  • (5) Iterative: use fine-tuned model to generate next round of chains, retrain.
  • Foundation of o1, DeepSeek-R1, and many modern reasoning models.
  • Test-time compute becomes tunable.
#llm#alignmentPermalink & quiz →

In RLHF, why does the reward model become unreliable as training progresses?

hard
  • Because the policy moves off the distribution the reward model was trained on.
  • Preference data came from earlier, weaker samples, so the reward model is accurate there and increasingly extrapolating as the policy improves, which the policy then exploits: it finds text scoring highly under the reward model that humans do not actually prefer.
  • This is why the KL penalty against the reference policy exists, since it keeps the policy inside the region where the reward model was fit rather than expressing a preference for the original model.
  • The operational answers are to monitor the KL divergence as a first-class metric, collect fresh preference data on current samples and retrain the reward model, and keep a held-out human evaluation as the ground truth, because reward going up is not evidence of quality going up.
#llm#alignmentPermalink & quiz →

Why do many teams choose DPO over PPO for alignment?

medium
  • Mostly engineering cost.
  • PPO-based RLHF requires four models in memory at once, the policy, the reference, the reward model and the critic, plus a generation loop inside training, which makes the pipeline expensive, slow and full of hyperparameters that can silently ruin a run.
  • DPO reformulates preference optimization so the reward model is implicit, turning the problem into a supervised loss on preference pairs, and it needs only the policy and a frozen reference.
  • That makes it dramatically simpler to implement, reproduce and debug.
  • The tradeoff is that DPO learns from a fixed preference set and cannot explore, so it does not benefit from on-policy samples, and at the frontier iterative or online variants that reintroduce fresh sampling tend to close the remaining quality gap.
#llm#alignmentPermalink & quiz →

Practise Reinforcement Learning